Этапы формирования цены

Определение способных повлиять на цену товара факторов, не зависящих от компании

Определение цели, с которой производится расчет цены на товар или услугу

Выбор способа формирования стоимости

Разработка стратегии, по которой будет определяться цена

Осуществление рыночной корректировки стоимости товара или услуги

План :

EDA

Прогнозирование временных рядов

План прогноза спроса

Эластичность на примере

План разработки приложения для ценообразования

EDA

import pandas as pd
import numpy as np

!ls

1510001_agr.csv       ZX10208_agr.csv       second_notebook.ipynb
1510002_agr.csv       conc_df.csv           sku_dict.csv
1510003_agr.csv       promo_1510002.csv

sku_dict= pd.read_csv('sku_dict.csv')

sku_dict.head()

	sku_id	fincode	ui1_code	ui2_code	ui3_code	vendor	brand_code	creation_date	expiration_date
0	43207	A1	A10	A1000	A1000001	A10001	A1000001	2018-01-01	2200-01-01 00:00:00
1	87209	UA	UA1	UA100	UA10000	BLW9FG	484JDFU5	2019-09-05	2200-01-01 00:00:00
2	33513	G9	G91	G9100	G910003	7I77VL	MMC2P4Y3	2018-07-21	2200-01-01 00:00:00
3	47505	AQ	AQ1	AQ100	AQ10000	4HJSP5	8FYENEWM	2018-03-26	2200-01-01 00:00:00
4	53350	6L	6L1	6L100	6L10005	JSOSD5	YTE7JS4H	2018-11-30	2200-01-01 00:00:00

sku_dict.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 10000 entries, 0 to 9999
Data columns (total 9 columns):
 #   Column           Non-Null Count  Dtype 
---  ------           --------------  ----- 
 0   sku_id           10000 non-null  int64 
 1   fincode          10000 non-null  object
 2   ui1_code         10000 non-null  object
 3   ui2_code         10000 non-null  object
 4   ui3_code         10000 non-null  object
 5   vendor           10000 non-null  object
 6   brand_code       10000 non-null  object
 7   creation_date    10000 non-null  object
 8   expiration_date  10000 non-null  object
dtypes: int64(1), object(8)
memory usage: 703.2+ KB

for col in sku_dict.columns:
    print(col,sku_dict['{}'.format(col)].nunique())

sku_id 10000
fincode 31
ui1_code 46
ui2_code 109
ui3_code 804
vendor 25
brand_code 63
creation_date 669
expiration_date 582

sku_dict.fincode.unique()

array(['A1', 'UA', 'G9', 'AQ', '6L', '4W', 'ZR', '15', 'SI', 'BY', '70',
       'BJ', 'RI', 'MB', 'YT', 'K9', 'FE', 'Z8', 'GY', 'M2', 'XG', 'LT',
       'EH', 'ZX', 'H7', '8J', 'AG', 'B1', 'Q5', 'GV', 'B0'], dtype=object)

sku_dict.ui1_code.unique()

array(['A10', 'UA1', 'G91', 'AQ1', '6L1', '4W1', 'ZR1', '151', 'SI1',
       'BY1', '701', 'BJ1', 'RI1', 'MB1', 'YT1', 'K91', 'FE1', 'Z81',
       'GY1', 'M21', 'XG1', 'LT1', 'EH1', 'ZX1', 'H71', '8J1', 'AG1',
       'B11', 'Q51', 'GV1', 'B01', 'EH0', '700', 'M20', 'K90', 'GV0',
       'ZX0', 'B00', 'G90', 'SI0', 'BJ0', 'RI0', 'UA0', 'YT0', 'Z80',
       '4W0'], dtype=object)

sku_dict.ui2_code.unique()

array(['A1000', 'UA100', 'G9100', 'AQ100', '6L100', '4W100', 'ZR101',
       '15100', 'SI100', 'BY100', '70102', 'BJ100', 'G9101', 'RI100',
       'MB100', 'YT102', 'MB101', 'K9102', 'FE100', 'Z8100', 'GY100',
       'UA102', 'M2100', 'XG100', 'LT102', '70101', 'EH100', 'BY101',
       'LT100', 'ZX101', 'H7100', 'XG102', '4W102', 'ZR100', '8J100',
       'AG100', 'AG101', 'K9100', 'Z8101', 'MB102', 'B1100', '6L102',
       'Q5100', 'GV102', 'B0100', 'AQ101', 'GV100', '70100', 'ZR102',
       'UA101', 'GV101', 'M2102', 'LT101', 'GY101', 'YT100', 'AG102',
       'H7101', '6L101', 'SI101', 'B1102', '8J102', 'RI101', 'B1101',
       'BJ102', 'RI102', 'AQ102', 'GY102', 'EH102', 'Q5102', 'EH101',
       '4W101', 'BJ101', 'BY102', 'FE101', 'G9102', 'M2101', '15101',
       'B0102', 'XG101', 'ZX100', 'SI102', 'B0101', '15102', '8J101',
       'K9101', 'Q5101', 'EH000', 'YT101', 'H7102', 'ZX102', 'FE102',
       'Z8102', '70000', 'M2000', 'K9000', 'EH002', 'GV000', 'ZX001',
       'B0000', 'G9001', 'SI000', 'BJ000', 'G9000', 'RI000', 'UA000',
       'YT001', 'Z8000', 'SI001', '4W000'], dtype=object)

sku_dict.ui3_code.unique()

array(['A1000001', 'UA10000', 'G910003', 'AQ10000', '6L10005', '4W10000',
       'ZR10100', 'AQ10008', '1510004', 'SI10006', 'AQ10005', 'BY10001',
       '7010200', 'AQ10001', 'BJ10001', 'G910100', 'RI10007', 'MB10006',
       'YT10204', 'MB10101', 'K910201', 'FE10002', 'Z810000', 'GY10000',
       'UA10200', 'M210008', 'UA10005', 'XG10002', 'LT10200', 'YT10200',
       '7010100', 'EH10000', 'BY10106', 'LT10007', 'ZX10102', 'H710000',
       'XG10207', 'XG10004', 'BY10000', '4W10207', 'ZR10000', '8J10004',
       'H710005', 'AG10002', 'Z810005', 'AG10100', 'K910005', 'UA10203',
       'EH10007', 'G910007', 'G910000', 'Z810100', '8J10003', 'ZR10005',
       'MB10201', 'LT10000', 'SI10005', 'B110000', '6L10207', 'BJ10000',
       'Q510000', 'GV10200', 'B010000', 'LT10207', 'AQ10105', 'EH10004',
       'M210000', 'GV10000', 'K910200', '7010000', 'ZR10208', 'UA10100',
       'ZX10104', 'GV10100', 'ZR10006', 'XG10000', '7010203', '1510007',
       'AG10000', '1510000', 'RI10000', 'LT10006', 'M210200', '6L10203',
       'LT10106', 'ZX10100', 'LT10208', '8J10000', 'AG10006', 'GY10100',
       'FE10004', 'RI10008', 'YT10000', 'MB10000', 'AG10200', 'FE10001',
       'H710100', '6L10100', 'LT10008', 'K910000', 'BY10105', '4W10006',
       'M210201', 'B110003', 'AQ10007', 'SI10105', 'FE10000', 'Q510007',
       'B110200', 'K910002', '8J10202', 'BJ10007', 'GY10003', 'B110002',
       'LT10206', 'RI10102', '4W10002', 'B110103', 'FE10008', 'AG10001',
       'SI10000', 'MB10102', 'BJ10200', 'GV10102', 'G910002', 'G910005',
       'GV10007', 'RI10200', '8J10002', 'RI10006', 'GY10106', 'AQ10206',
       'GY10200', 'SI10007', '6L10208', 'EH10200', 'Q510200', 'EH10100',
       '6L10000', '4W10100', 'EH10008', 'BJ10100', 'BY10200', 'H710007',
       'M210006', 'AG10005', 'GV10006', 'AG10004', 'FE10100', 'AG10105',
       'YT10003', 'G910200', 'SI10003', 'MB10005', 'FE10006', 'AQ10004',
       'LT10100', 'B010002', 'XG10203', '8J10200', 'M210100', 'EH10106',
       'SI10100', 'H710001', 'EH10101', 'AQ10205', '7010201', '8J10008',
       'ZR10001', '7010005', 'Q510207', 'XG10001', 'B110008', '1510100',
       'B010200', 'XG10102', 'M210002', 'XG10206', '1510006', 'K910006',
       'UA10205', 'ZX10005', 'SI10108', 'Q510006', 'H710003', 'RI10002',
       'SI10200', 'EH10003', 'M210207', 'MB10200', 'B010101', '4W10001',
       '8J10005', 'B010100', 'K910007', '7010001', 'MB10002', 'XG10003',
       'ZX10002', 'XG10006', '7010004', 'B110004', 'G910008', 'ZX10000',
       'SI10008', '1510200', 'ZR10200', 'UA10007', 'Q510003', '1510105',
       '4W10004', 'ZR10008', '4W10005', 'BJ10002', 'GY10007', 'XG10106',
       '4W10205', 'RI10004', 'Z810006', '8J10104', '4W10200', 'Q510004',
       'RI10100', 'RI10203', 'K910104', 'GV10001', 'MB10003', 'K910008',
       '6L10006', 'ZX10006', 'BY10007', '8J10100', '1510001', 'AG10107',
       'B110100', 'SI10002', 'GY10105', 'BJ10104', 'SI10004', 'AQ10003',
       'EH10202', 'Q510100', 'EH10002', '8J10007', 'GV10005', 'B010106',
       'Q510005', 'B010005', 'LT10004', 'ZX10008', 'EH00004', 'UA10006',
       'RI10206', 'YT10004', '8J10001', 'GY10108', 'BY10004', 'M210208',
       'YT10107', 'BY10005', 'BJ10008', 'Z810007', '6L10200', 'Q510205',
       'UA10008', 'AQ10006', '1510002', 'YT10100', 'K910100', '7010106',
       'UA10004', 'M210003', 'RI10106', 'AQ10100', 'AQ10208', 'H710200',
       'AG10007', 'ZX10205', 'GV10008', 'K910103', 'LT10201', 'GY10002',
       'M210102', '7010102', 'BJ10005', 'M210007', '7010208', 'ZR10203',
       'RI10001', 'G910203', '8J10006', 'H710002', '7010007', '6L10105',
       'GV10106', 'LT10002', 'EH10001', '4W10003', 'B010007', 'YT10102',
       '7010107', 'K910003', 'GY10004', 'EH10006', '7010207', 'BY10204',
       'YT10008', 'BY10100', '4W10102', 'FE10207', 'K910108', 'FE10202',
       'AG10003', 'Z810003', 'ZR10102', 'ZX10007', 'LT10105', 'XG10007',
       'XG10100', '4W10101', 'FE10200', '1510003', 'GV10205', 'MB10207',
       '7010002', 'K910001', 'ZR10004', '6L10004', 'SI10106', 'B110102',
       'B010001', 'YT10104', 'UA10002', '4W10008', 'BY10006', 'XG10200',
       'Z810200', 'MB10008', '6L10007', 'YT10101', 'M210001', 'FE10003',
       'AQ10108', 'XG10204', 'SI10001', 'AG10204', 'BY10003', 'AG10202',
       'YT10006', 'FE10203', '6L10001', 'Z810002', 'G910104', 'GY10008',
       'GV10201', 'H710107', '6L10008', 'GY10001', '1510103', 'K910206',
       'FE10208', '8J10204', 'G910006', 'GY10005', 'RI10005', 'XG10103',
       'Z810008', 'EH10005', 'BY10208', 'ZR10202', 'MB10106', 'ZX10204',
       'GV10103', 'AQ10200', 'UA10106', 'YT10005', 'Z810105', '6L10205',
       'BY10008', 'BY10102', 'MB10007', 'B010201', 'G910208', 'B110205',
       'XG10202', 'GV10002', 'B010008', 'XG10105', 'M210105', 'BJ10006',
       'UA10003', 'FE10005', 'LT10203', 'M210005', 'ZR10104', 'AG10207',
       'G910107', 'FE10107', 'B110007', 'ZX10200', 'BJ10202', 'Q510108',
       'MB10001', 'RI10204', 'GY10006', '6L10002', '4W10105', 'B110001',
       'ZR10107', 'GV10104', '8J10101', 'BY10002', 'B110206', 'ZX10001',
       'XG10005', 'GY10107', 'LT10001', 'B110207', 'H710203', 'MB10105',
       'YT10002', 'YT10205', 'GY10204', 'FE10204', 'G910004', 'RI10104',
       'AG10008', 'LT10005', '1510206', 'YT10202', 'XG10205', 'ZR10002',
       '7000005', 'FE10105', 'B010003', 'ZX10004', 'GY10205', 'Q510001',
       '1510201', 'K910004', 'XG10008', 'B010204', 'UA10208', 'ZR10106',
       'ZX10203', 'H710206', '8J10102', 'BJ10204', 'ZR10007', 'B010103',
       'AG10208', 'FE10108', 'SI10101', 'H710205', 'B010205', 'MB10100',
       'Z810004', 'ZX10208', 'BJ10003', 'K910202', '8J10203', '7010006',
       'BJ10004', 'G910201', '6L10103', 'M210106', 'UA10207', 'H710103',
       'XG10201', 'UA10102', '7010206', 'B010206', 'M210004', '4W10106',
       'FE10106', '1510008', 'BY10201', '7010204', '7010003', '8J10207',
       'Q510204', 'MB10203', 'Q510103', 'LT10108', 'M210103', '6L10003',
       'H710207', 'Z810001', 'EH10102', 'ZR10101', 'G910204', '4W10007',
       '4W10206', 'GY10203', 'LT10107', 'SI10206', 'G910103', 'Q510106',
       'GV10004', '6L10202', 'YT10106', 'XG10107', 'K910106', 'UA10107',
       'YT10203', 'YT10001', '7010108', 'Q510206', 'GV10003', 'EH10206',
       'GV10105', 'M210206', 'GY10202', 'G910001', 'Z810201', 'Z810207',
       'UA10202', 'FE10206', 'Q510008', 'M200004', 'UA10108', 'EH10201',
       'H710008', 'Z810108', 'BJ10106', 'RI10003', 'FE10102', '6L10101',
       'AG10206', 'ZR10205', 'BJ10201', 'MB10202', 'EH10207', 'BY10203',
       'ZR10201', 'B010107', 'ZX10105', 'BY10205', '7010008', '4W10202',
       'G910102', 'BY10101', 'K900002', 'Z810102', '7010101', 'H710104',
       'Q510104', '4W10208', 'YT10207', 'FE10007', 'MB10004', 'EH10108',
       'LT10003', '4W10104', 'YT10007', '1510005', 'AQ10203', 'B010004',
       'EH00200', 'GV10101', 'M210101', 'YT10105', 'BJ10207', 'GY10201',
       'B010105', 'UA10104', '8J10206', 'ZX10003', '7010104', 'B110006',
       'Q510002', '1510108', 'K910204', 'GV10203', 'UA10001', 'AQ10201',
       '4W10203', 'ZX10202', 'B110005', 'G910205', '4W10204', 'MB10206',
       'GV00000', 'BJ10102', 'B010102', 'SI10202', 'MB10205', 'GY10103',
       'SI10201', 'ZR10204', 'G910202', 'H710006', 'GY10101', 'FE10101',
       'ZR10003', 'XG10108', 'EH00000', 'AG10108', '4W10108', '6L10107',
       'AQ10107', 'Z810103', 'RI10208', 'Q510202', 'SI10107', 'ZR10206',
       'SI10207', 'RI10105', 'K910208', 'M210107', 'EH10204', 'M210203',
       'YT10103', 'H710004', 'GV10202', '6L10106', 'B010207', 'SI10203',
       'EH10103', 'RI10201', 'ZX00100', 'B000000', 'GY10208', 'BY10108',
       'SI10102', 'ZR10103', 'XG10208', 'YT10201', 'ZX10103', 'YT10108',
       'MB10107', 'EH10205', '8J10201', 'BJ10205', 'LT10205', '1510203',
       '4W10201', '8J10208', 'GV10206', 'UA10103', '7010205', 'ZX10108',
       'Q510105', 'RI10202', 'LT10202', 'M210202', '4W10103', 'G900107',
       'MB10104', 'M210204', 'AQ10202', 'GV10108', 'H710106', 'AQ10204',
       'LT10204', 'XG10101', 'AQ10106', '6L10102', 'M210205', 'FE10103',
       'B010108', 'LT10103', 'BJ10101', 'Z810104', 'SI10104', 'ZR10105',
       'B010208', 'MB10204', 'BY10107', 'BY10202', 'GY10102', 'MB10108',
       'EH10104', 'B110106', 'RI10101', 'B010006', '6L10204', '6L10201',
       'AG10205', 'SI10103', 'K910207', 'AQ10102', '7010105', 'H710202',
       'B110107', 'AQ10002', 'BY10103', '8J10107', '1510208', '1510102',
       'BY10207', 'SI00000', 'GY10104', 'Z810202', 'LT10101', '8J10108',
       'G910206', 'GY10207', '6L10206', 'ZX10107', 'B110101', 'H710201',
       'BJ10208', '4W10107', '6L10108', 'ZX10106', 'AQ10103', '7010202',
       'Q510102', 'YT10208', 'SI10204', 'Z810204', 'M210108', '1510104',
       'AG10103', 'G910207', 'K910105', 'GV10208', 'ZX10101', 'BJ10108',
       'H710108', 'G910101', '6L10104', 'AQ10104', 'UA10101', 'EH10203',
       'GY10206', 'MB10103', 'BJ00005', 'B110201', '8J10106', 'BJ10206',
       'B110108', 'RI10108', 'K910205', 'GV00007', 'UA10105', 'Z810208',
       'EH10208', 'G900006', '1510107', 'FE10205', 'K910102', '7010103',
       'SI10205', 'UA10201', 'LT10102', 'AQ10101', 'ZR10207', 'G910105',
       'G910106', 'BJ10203', 'RI00004', 'B010104', 'UA10204', 'UA00000',
       'SI10208', 'K910101', 'AG10104', '8J10105', 'BJ10107', 'ZR10108',
       'Q510101', 'H710204', 'B110203', 'ZX10207', 'Z810206', 'LT10104',
       'RI10103', 'RI10107', 'Z810101', '1510207', '1510205', 'Q510208',
       'H710208', 'RI10205', 'GV10204', 'BJ10105', '8J10205', '1510106',
       'BY10104', 'Q510107', 'GV10207', 'H710105', 'AG10101', 'ZX10206',
       'BJ10103', 'XG10104', 'M210104', 'AQ10207', 'UA10206', 'M200003',
       'YT00100', 'Z800004', 'RI10207', 'SI00100', 'GV10107', 'K910107',
       'ZX10201', 'Z810107', 'BY10206', '4W00008', 'AG10106', 'G900000'],
      dtype=object)

sku_dict.brand_code.value_counts()

4WL34OA2    459
8FYENEWM    439
THSTSG81    384
02Q78ADY    304
MMC2P4Y3    294
           ... 
SKXXW8V9     59
JK4UKBI2     52
74CKJ9XQ     51
H1BWFJRN     33
A1000001      1
Name: brand_code, Length: 63, dtype: int64

sku_dict.vendor.value_counts()

UEYMBB    458
TJYT3A    447
71SM2O    445
7I77VL    440
LDLJFP    435
T9BUA7    434
2CI2A8    432
UH2WWF    430
SJP384    425
AWD3XQ    425
4HJSP5    419
BLW9FG    418
B4TA32    411
3GFIMK    411
SHL570    408
AO59EQ    403
VWHZMG    400
VZ44WY    399
H2OK0Z    399
FI3ASO    398
JSOSD5    398
GVL335    391
4O4WSH    390
CHQKSO    383
A10001      1
Name: vendor, dtype: int64

sku_dict.groupby(['fincode', 'ui1_code']).sku_id.agg('count').reset_index().sort_values('fincode')

	fincode	ui1_code	sku_id
0	15	151	326
1	4W	4W0	1
2	4W	4W1	345
3	6L	6L1	305
4	70	700	1
5	70	701	305
6	8J	8J1	322
7	A1	A10	1
8	AG	AG1	309
9	AQ	AQ1	347
11	B0	B01	320
10	B0	B00	1
12	B1	B11	293
13	BJ	BJ0	1
14	BJ	BJ1	341
15	BY	BY1	316
16	EH	EH0	3
17	EH	EH1	310
18	FE	FE1	299
19	G9	G90	3
20	G9	G91	336
22	GV	GV1	344
21	GV	GV0	2
23	GY	GY1	619
24	H7	H71	327
25	K9	K90	1
26	K9	K91	326
27	LT	LT1	347
28	M2	M20	2
29	M2	M21	311
30	MB	MB1	305
31	Q5	Q51	337
33	RI	RI1	328
32	RI	RI0	1
34	SI	SI0	2
35	SI	SI1	314
36	UA	UA0	1
37	UA	UA1	344
38	XG	XG1	343
39	YT	YT0	1
40	YT	YT1	332
41	Z8	Z80	1
42	Z8	Z81	293
43	ZR	ZR1	326
44	ZX	ZX0	1
45	ZX	ZX1	307

sku_dict.groupby(['ui1_code', 'ui3_code']).sku_id.agg('count').reset_index().sort_values('ui1_code')

	ui1_code	ui3_code	sku_id
0	151	1510000	123
23	151	1510208	4
22	151	1510207	2
21	151	1510206	2
20	151	1510205	1
...	...	...	...
778	ZX1	ZX10001	11
777	ZX1	ZX10000	112
802	ZX1	ZX10207	1
789	ZX1	ZX10103	1
803	ZX1	ZX10208	2

804 rows × 3 columns

ZX10208 Group

ZX10208_df = pd.read_csv('ZX10208_agr.csv')

ZX10208_df.head()

	dates	SKU	price_per_sku	num_purchases
0	2018-04-20	59567	169.19	3
1	2018-04-24	59567	171.71	3
2	2018-04-27	59567	174.04	3
3	2018-04-30	59567	NaN	1
4	2018-05-03	59567	176.38	2

ZX10208_df.SKU.unique()

array([59567, 82947])

sku_dict[sku_dict.sku_id.isin(ZX10208_df.SKU.unique())]

	sku_id	fincode	ui1_code	ui2_code	ui3_code	vendor	brand_code	creation_date	expiration_date
1484	59567	ZX	ZX1	ZX102	ZX10208	SJP384	3FAIRJ4W	2018-04-16	2019-04-21 00:00:00
4048	82947	ZX	ZX1	ZX102	ZX10208	TJYT3A	3FAIRJ4W	2018-09-26	2200-01-01 00:00:00

ZX10208_df.dates = pd.to_datetime(ZX10208_df.dates)

ZX10208_df[ZX10208_df.SKU == 82947].dates.describe()

/Users/sgafarov/opt/anaconda3/lib/python3.7/site-packages/ipykernel_launcher.py:1: FutureWarning: Treating datetime data as categorical rather than numeric in `.describe` is deprecated and will be removed in a future version of pandas. Specify `datetime_is_numeric=True` to silence this warning and adopt the future behavior now.
  """Entry point for launching an IPython kernel.

count                     103
unique                    103
top       2019-09-09 00:00:00
freq                        1
first     2018-09-28 00:00:00
last      2019-12-28 00:00:00
Name: dates, dtype: object

#import visualization libraries
import matplotlib.pyplot as plt
import seaborn as sns
color = sns.color_palette()

%matplotlib inline

plt.figure(figsize=(12,8))
plt.scatter(x='dates', y='price_per_sku', data=ZX10208_df)
plt.ylabel('price_per_sku', fontsize=12)
plt.xlabel('dates', fontsize=12)
plt.xticks(rotation='vertical')
plt.show()

# week month and year
ZX10208_df['week_num'] = ZX10208_df.dates.dt.week
ZX10208_df['year'] = ZX10208_df.dates.dt.year
ZX10208_df['months'] = ZX10208_df.dates.dt.month
ZX10208_df['year_week'] =  ZX10208_df['year'].astype(str) + '_' + ZX10208_df['week_num'].astype(str)

/Users/sgafarov/opt/anaconda3/lib/python3.7/site-packages/ipykernel_launcher.py:2: FutureWarning: Series.dt.weekofyear and Series.dt.week have been deprecated.  Please use Series.dt.isocalendar().week instead.
  

ZX10208_df.head()

	dates	SKU	price_per_sku	num_purchases	week_num	year	months	year_week
0	2018-04-20	59567	169.19	3	16	2018	4	2018_16
1	2018-04-24	59567	171.71	3	17	2018	4	2018_17
2	2018-04-27	59567	174.04	3	17	2018	4	2018_17
3	2018-04-30	59567	NaN	1	18	2018	4	2018_18
4	2018-05-03	59567	176.38	2	18	2018	5	2018_18

ZX10208_gr = ZX10208_df.groupby(['year_week','SKU']).price_per_sku.mean().reset_index()

ZX10208_gr = ZX10208_gr.rename(columns={'price_per_sku':'mean_per_week'})

ZX10208_df = ZX10208_df.merge(ZX10208_gr, how='left', on=['year_week','SKU'])

ZX10208_df

	dates	SKU	price_per_sku	num_purchases	week_num	year	months	year_week	mean_per_week
0	2018-04-20	59567	169.19	3	16	2018	4	2018_16	169.190
1	2018-04-24	59567	171.71	3	17	2018	4	2018_17	172.875
2	2018-04-27	59567	174.04	3	17	2018	4	2018_17	172.875
3	2018-04-30	59567	NaN	1	18	2018	4	2018_18	176.435
4	2018-05-03	59567	176.38	2	18	2018	5	2018_18	176.435
...	...	...	...	...	...	...	...	...	...
254	2019-12-23	82947	NaN	1	52	2019	12	2019_52	194.810
255	2019-12-24	82947	195.46	2	52	2019	12	2019_52	194.810
256	2019-12-25	82947	195.21	1	52	2019	12	2019_52	194.810
257	2019-12-26	82947	193.79	1	52	2019	12	2019_52	194.810
258	2019-12-28	82947	194.78	1	52	2019	12	2019_52	194.810

259 rows × 9 columns

plt.figure(figsize=(12,8))
plt.scatter(x='year_week', y='mean_per_week', data=ZX10208_df[ZX10208_df.SKU==59567].sort_values('dates'))
plt.ylabel('price mean_per_week', fontsize=12)
plt.xlabel('dates', fontsize=12)
plt.xticks(rotation='vertical')
plt.show()

pivot_t = ZX10208_df.pivot_table(index='dates',columns='SKU', values='price_per_sku')

pivot_t

SKU	59567	82947
dates
2018-04-20	169.19	NaN
2018-04-24	171.71	NaN
2018-04-27	174.04	NaN
2018-05-03	176.38	NaN
2018-05-04	176.49	NaN
...	...	...
2019-12-21	NaN	195.96
2019-12-24	NaN	195.46
2019-12-25	NaN	195.21
2019-12-26	NaN	193.79
2019-12-28	NaN	194.78

181 rows × 2 columns

plt.figure(figsize=(12,8))
sns.lineplot(data =pivot_t)

<matplotlib.axes._subplots.AxesSubplot at 0x7fcc56cd5510>

1510001 group

df_1510001 = pd.read_csv('1510001_agr.csv')

df_1510001

	dates	SKU	price_per_sku	num_purchases
0	2018-08-02	39665	85.940	1
1	2018-08-03	39665	85.940	1
2	2018-08-04	39665	85.940	1
3	2018-08-05	39665	85.940	1
4	2018-08-06	39665	86.038	1
...	...	...	...	...
3935	2019-12-25	36108	80.094	2
3936	2019-12-28	36108	80.094	2
3937	2019-12-29	36108	80.094	1
3938	2019-12-30	36108	76.205	1
3939	2019-12-31	36108	76.205	1

3940 rows × 4 columns

df_1510001.dates = pd.to_datetime(df_1510001.dates)

pivot_t = df_1510001.pivot_table(index='dates',columns='SKU',values='price_per_sku')

pivot_t

SKU	27240	27885	34130	36108	39665	40194	45164	48671	49596	59631	63995	70252	70315	72449
dates
2018-03-02	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	107.784	NaN	NaN
2018-03-03	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	107.784	NaN	NaN
2018-03-04	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	107.784	NaN	NaN
2018-03-05	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	107.882	NaN	NaN
2018-03-06	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	107.882	NaN	NaN
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
2019-12-27	128.652	NaN	104.536	NaN	95.978	135.931	82.652	152.815	148.536	NaN	138.536	NaN	71.931	156.257
2019-12-28	NaN	NaN	NaN	80.094	95.978	NaN	82.652	152.815	NaN	108.536	138.536	117.815	71.931	156.257
2019-12-29	128.652	NaN	104.536	80.094	95.978	135.931	82.652	NaN	148.536	NaN	138.536	NaN	71.931	156.257
2019-12-30	NaN	NaN	NaN	76.205	87.015	133.475	79.840	148.205	NaN	NaN	132.110	NaN	68.015	NaN
2019-12-31	NaN	NaN	98.840	76.205	87.015	133.475	NaN	148.205	142.475	NaN	132.110	112.475	68.015	149.110

650 rows × 14 columns

plt.figure(figsize=(12,8))
sns.lineplot(data=pivot_t,dashes=False)

<matplotlib.axes._subplots.AxesSubplot at 0x7fcc5726dfd0>

df_1510001.SKU.nunique()

14

df_1510001.SKU.unique()

array([39665, 63995, 70252, 34130, 70315, 27885, 59631, 48671, 72449,
       49596, 27240, 45164, 40194, 36108])

df_1510001[df_1510001.SKU == 39665].num_purchases.describe()

count    354.000000
mean       1.519774
std        0.694702
min        1.000000
25%        1.000000
50%        1.000000
75%        2.000000
max        3.000000
Name: num_purchases, dtype: float64

plt.figure(figsize=(12,8))

sns.lineplot(x =df_1510001[df_1510001.SKU == 39665].num_purchases , y = df_1510001[df_1510001.SKU == 39665].price_per_sku)

<matplotlib.axes._subplots.AxesSubplot at 0x7fcc52723fd0>

1510002 group

df_1510002 = pd.read_csv('1510002_agr.csv')

df_1510002.dates = pd.to_datetime(df_1510002.dates)

pivot_t = df_1510002.pivot_table(index='dates',columns='SKU',values='num_purchases')

pivot_t

SKU	10334	17812	36096	38767	43678	57561	64619	68211	69997	72784	83329	90611	98255
dates
2018-01-24	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	9.0	NaN	NaN	NaN
2018-01-25	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	9.0	NaN	NaN	NaN
2018-01-27	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	7.0	NaN	NaN	NaN
2018-01-28	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	7.0	NaN	NaN	NaN
2018-01-29	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	9.0	NaN	NaN	NaN
...	...	...	...	...	...	...	...	...	...	...	...	...	...
2019-12-27	9.0	3.0	13.0	NaN	9.0	NaN	4.0	NaN	NaN	7.0	9.0	NaN	NaN
2019-12-28	9.0	4.0	13.0	8.0	10.0	NaN	NaN	NaN	NaN	NaN	NaN	4.0	NaN
2019-12-29	8.0	NaN	14.0	9.0	NaN	NaN	4.0	NaN	NaN	NaN	9.0	5.0	NaN
2019-12-30	8.0	3.0	NaN	8.0	9.0	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
2019-12-31	9.0	3.0	NaN	8.0	9.0	NaN	NaN	NaN	NaN	NaN	10.0	6.0	NaN

702 rows × 13 columns

plt.figure(figsize=(12,8))
sns.lineplot(data=pivot_t,dashes=False)

<matplotlib.axes._subplots.AxesSubplot at 0x7fcc56af5c50>

promo_df = pd.read_csv('promo_1510002.csv')

promo_df

	year	week_num	SKU	discount
0	2019	1	90611	0.10
1	2019	2	10334	0.10
2	2019	3	68211	0.10
3	2019	4	10334	0.10
4	2019	5	72784	0.20
5	2019	6	68211	0.10
6	2019	7	38767	0.10
7	2019	8	83329	0.10
8	2019	9	83329	0.05
9	2019	10	90611	0.10
10	2019	11	57561	0.05
11	2019	12	72784	0.10
12	2019	13	38767	0.05
13	2019	14	57561	0.10
14	2019	15	57561	0.20
15	2019	16	90611	0.05
16	2019	17	43678	0.20
17	2019	18	10334	0.20
18	2019	19	38767	0.05
19	2019	20	17812	0.10
20	2019	21	38767	0.05
21	2019	22	57561	0.05
22	2019	23	10334	0.05
23	2019	24	69997	0.20
24	2019	25	17812	0.10
25	2019	26	17812	0.10
26	2019	27	17812	0.05
27	2019	28	10334	0.10
28	2019	29	90611	0.05
29	2019	30	10334	0.20
30	2019	31	69997	0.20
31	2019	32	57561	0.10
32	2019	33	43678	0.10
33	2019	34	17812	0.20
34	2019	35	90611	0.20
35	2019	36	57561	0.10
36	2019	37	43678	0.10
37	2019	38	57561	0.05
38	2019	39	36096	0.10
39	2019	40	72784	0.10
40	2019	41	64619	0.05
41	2019	42	38767	0.05
42	2019	43	17812	0.05
43	2019	44	43678	0.10
44	2019	45	17812	0.10
45	2019	46	83329	0.10
46	2019	47	90611	0.20
47	2019	48	36096	0.05
48	2019	49	10334	0.05
49	2019	50	10334	0.05
50	2019	51	17812	0.10
51	2019	52	36096	0.10

promo_df.discount = 1 - promo_df.discount

df_1510002['year'] = df_1510002.dates.dt.year
df_1510002['week_num'] = df_1510002.dates.dt.week
df_1510002 = df_1510002[df_1510002.dates >= pd.to_datetime('2019-01-01')]

/Users/sgafarov/opt/anaconda3/lib/python3.7/site-packages/ipykernel_launcher.py:2: FutureWarning: Series.dt.weekofyear and Series.dt.week have been deprecated.  Please use Series.dt.isocalendar().week instead.
  

df_72784 =  df_1510002[df_1510002.SKU == 72784]

df_72784.groupby(['year','week_num']).num_purchases.max().reset_index()

	year	week_num	num_purchases
0	2019	1	8
1	2019	2	7
2	2019	3	7
3	2019	4	7
4	2019	5	11
5	2019	6	6
6	2019	7	6
7	2019	8	8
8	2019	9	8
9	2019	10	8
10	2019	11	8
11	2019	12	10
12	2019	13	6
13	2019	14	7
14	2019	15	6
15	2019	16	7
16	2019	17	8
17	2019	18	7
18	2019	19	6
19	2019	20	6
20	2019	21	8
21	2019	22	7
22	2019	23	8
23	2019	24	7
24	2019	25	8
25	2019	26	7
26	2019	27	7
27	2019	28	7
28	2019	29	6
29	2019	30	8
30	2019	31	6
31	2019	32	8
32	2019	33	8
33	2019	34	7
34	2019	35	7
35	2019	36	8
36	2019	37	8
37	2019	38	8
38	2019	39	7
39	2019	40	10
40	2019	41	8
41	2019	42	8
42	2019	43	6
43	2019	44	7
44	2019	45	8
45	2019	46	8
46	2019	47	8
47	2019	48	7
48	2019	49	8
49	2019	50	7
50	2019	51	6
51	2019	52	7

df_1510002 = df_1510002.merge(promo_df, how='left', on= ['SKU', 'year', 'week_num'])

df_1510002['real_price_per_sku'] = df_1510002.price_per_sku * (1+df_1510002.discount)

df_1510002[~df_1510002.discount.isnull()]

	dates	SKU	price_per_sku	num_purchases	year	week_num	discount	real_price_per_sku
0	2019-01-01	90611	79.5420	7	2019	1	0.1	87.49620
13	2019-01-02	90611	79.5420	7	2019	1	0.1	87.49620
20	2019-01-03	90611	79.5420	9	2019	1	0.1	87.49620
27	2019-01-04	90611	79.5420	6	2019	1	0.1	87.49620
35	2019-01-05	90611	79.5420	6	2019	1	0.1	87.49620
...	...	...	...	...	...	...	...	...
2285	2019-12-26	36096	119.3868	14	2019	52	0.1	131.32548
2298	2019-12-27	36096	119.3868	13	2019	52	0.1	131.32548
2299	2019-12-28	36096	119.3868	13	2019	52	0.1	131.32548
2305	2019-12-29	36096	119.3868	14	2019	52	0.1	131.32548
2315	2019-12-31	90611	79.5420	6	2019	1	0.1	87.49620

281 rows × 8 columns

1510003 group

df_1 = pd.read_csv('1510003_agr.csv')

df_1

	dates	SKU	price_per_sku	num_purchases
0	2018-11-17	54404	80.095	11
1	2018-11-18	54404	80.095	12
2	2018-11-19	54404	81.796	10
3	2018-11-20	54404	81.796	11
4	2018-11-21	54404	81.796	10
...	...	...	...	...
2846	2019-12-27	49661	165.978	10
2847	2019-12-28	49661	165.978	9
2848	2019-12-29	49661	165.978	10
2849	2019-12-30	49661	159.205	8
2850	2019-12-31	49661	159.205	8

2851 rows × 4 columns

conc_df = pd.read_csv('conc_df.csv')

conc_df

	SKU	year	week_num	ret_net_code	price_per_sku
0	84395	2019	18	1	164.69400
1	84395	2019	18	2	164.69400
2	84395	2019	18	3	134.50010
3	84395	2019	22	1	165.19800
4	84395	2019	22	2	136.28835
...	...	...	...	...	...
3004	89272	2019	18	2	133.93820
3005	89272	2019	18	3	121.76200
3006	80308	2019	47	1	115.77984
3007	80308	2019	47	2	118.19192
3008	80308	2019	47	3	116.98588

3009 rows × 5 columns

df_1[df_1.SKU.isin(conc_df.SKU.unique())]

	dates	SKU	price_per_sku	num_purchases
0	2018-11-17	54404	80.095	11
1	2018-11-18	54404	80.095	12
2	2018-11-19	54404	81.796	10
3	2018-11-20	54404	81.796	11
4	2018-11-21	54404	81.796	10
...	...	...	...	...
2846	2019-12-27	49661	165.978	10
2847	2019-12-28	49661	165.978	9
2848	2019-12-29	49661	165.978	10
2849	2019-12-30	49661	159.205	8
2850	2019-12-31	49661	159.205	8

2851 rows × 4 columns

df_1.dates = pd.to_datetime(df_1.dates)

df_1['year'] = df_1.dates.dt.year
df_1['week_num'] = df_1.dates.dt.week

/Users/sgafarov/opt/anaconda3/lib/python3.7/site-packages/ipykernel_launcher.py:2: FutureWarning: Series.dt.weekofyear and Series.dt.week have been deprecated.  Please use Series.dt.isocalendar().week instead.
  

conc_aggr = conc_df.groupby(['SKU', 'year', 'week_num']).price_per_sku.agg(['min', 'max']).reset_index()

df_1 = df_1.merge(conc_aggr, how='left', on=['SKU','year', 'week_num'])

df_80308 = df_1[df_1.SKU == 80308]

df_80308

	dates	SKU	price_per_sku	num_purchases	year	week_num	min	max
853	2018-07-17	80308	113.940	8	2018	29	NaN	NaN
854	2018-07-18	80308	113.940	8	2018	29	NaN	NaN
855	2018-07-20	80308	113.940	8	2018	29	NaN	NaN
856	2018-07-21	80308	113.940	7	2018	29	NaN	NaN
857	2018-07-22	80308	113.940	8	2018	29	NaN	NaN
...	...	...	...	...	...	...	...	...
1235	2019-12-27	80308	122.536	10	2019	52	122.536	147.0432
1236	2019-12-28	80308	122.536	9	2019	52	122.536	147.0432
1237	2019-12-29	80308	122.536	9	2019	52	122.536	147.0432
1238	2019-12-30	80308	116.475	7	2019	1	111.816	128.1225
1239	2019-12-31	80308	116.475	5	2019	1	111.816	128.1225

387 rows × 8 columns

plt.figure(figsize=(12,8))
plt.scatter(x = df_80308[df_80308.price_per_sku >= df_80308['min']].dates, y =df_80308[df_80308.price_per_sku >= df_80308['min']].num_purchases)
plt.scatter(x =df_80308[df_80308.price_per_sku <= df_80308['min']].dates, y=df_80308[df_80308.price_per_sku <= df_80308['min']].num_purchases)

<matplotlib.collections.PathCollection at 0x7fcc5769ff10>

Прогнозирование временных рядов

from statsmodels.tsa.ar_model import AutoReg
from statsmodels.tsa.arima.model import ARIMA
from statsmodels.tsa.statespace.sarimax import SARIMAX
from statsmodels.tsa.vector_ar.var_model import VAR
from statsmodels.tsa.statespace.varmax import VARMAX
from statsmodels.tsa.holtwinters import SimpleExpSmoothing
from pandas.plotting import autocorrelation_plot
import warnings
warnings.filterwarnings('ignore')

ZX_df = pd.read_csv('ZX10208_agr.csv')
ZX_df['dates'] = pd.to_datetime(ZX_df.dates)
pivot_t = ZX_df.pivot_table(index='dates',columns='SKU', values='price_per_sku')

plt.figure(figsize=(12,8))
sns.lineplot(data =pivot_t)

<matplotlib.axes._subplots.AxesSubplot at 0x7fcc59408850>

ZX_df['price_per_sku'] = ZX_df.groupby('SKU').price_per_sku.ffill()

df_59567 = ZX_df[ZX_df.SKU == 59567]

train_df = df_59567[df_59567.dates <= pd.to_datetime('2019-04-01')]
test_df = df_59567[df_59567.dates > pd.to_datetime('2019-04-01')]

print(train_df.shape)
print(test_df.shape)

(145, 4)
(11, 10)

plt.figure(figsize=(12,8))
autocorrelation_plot(df_59567.price_per_sku.values)

<matplotlib.axes._subplots.AxesSubplot at 0x7fcc59382850>

#1 Autoregression
AR = AutoReg(train_df.price_per_sku.values, lags=20)
AR_model = AR.fit()

AR_model.params

array([ 1.60820207e+01,  1.06938563e+00, -1.46162089e-01, -1.30899380e-01,
        1.57571721e-01,  6.47218415e-02, -1.11758222e-01,  5.59600227e-02,
       -3.76933446e-02,  1.47311386e-02, -6.18697944e-02,  2.73159049e-01,
       -3.18091136e-01,  1.60588852e-01, -7.09107645e-02, -6.92044450e-02,
        1.41846291e-01, -1.33862976e-01,  4.72945316e-02,  4.09616542e-02,
       -3.34062939e-02])

AR_model.predict(start =len(train_df), end= len(train_df) + len(test_df)-1, 
                 dynamic=False)

array([180.44857402, 180.4443652 , 180.87926656, 181.70071796,
       180.70814628, 181.08981343, 181.11552473, 181.10372658,
       182.38490989, 182.0267328 , 182.55686698])

test_df['AR_model_prediction'] = AR_model.predict(start =len(train_df), end= len(train_df) + len(test_df)-1, 
                 dynamic=False)

test_df

	dates	SKU	price_per_sku	num_purchases	AR_model_prediction
192	2019-04-02	59567	181.86	2	180.448574
194	2019-04-06	59567	181.83	2	180.444365
195	2019-04-08	59567	181.83	1	180.879267
196	2019-04-09	59567	181.67	1	181.700718
197	2019-04-10	59567	180.08	2	180.708146
198	2019-04-11	59567	179.97	2	181.089813
199	2019-04-12	59567	179.03	1	181.115525
200	2019-04-13	59567	179.36	1	181.103727
202	2019-04-16	59567	178.61	2	182.384910
203	2019-04-18	59567	177.77	1	182.026733
204	2019-04-21	59567	177.77	1	182.556867

# Moving Avarage
MA = ARIMA(train_df.price_per_sku.values, order=(0,0,20))
MA_model = MA.fit()

MA_model.predict(start =len(train_df), end= len(train_df) + len(test_df)-1)

array([181.1411973 , 180.97153481, 181.78479433, 182.45938971,
       181.10404348, 181.8960975 , 181.05864524, 181.7544814 ,
       182.56178024, 181.3105422 , 182.60947083])

test_df['MA_model_prediction'] = MA_model.predict(start =len(train_df), end= len(train_df) + len(test_df)-1)

test_df

	dates	SKU	price_per_sku	num_purchases	AR_model_prediction	MA_model_prediction
192	2019-04-02	59567	181.86	2	180.448574	181.141197
194	2019-04-06	59567	181.83	2	180.444365	180.971535
195	2019-04-08	59567	181.83	1	180.879267	181.784794
196	2019-04-09	59567	181.67	1	181.700718	182.459390
197	2019-04-10	59567	180.08	2	180.708146	181.104043
198	2019-04-11	59567	179.97	2	181.089813	181.896098
199	2019-04-12	59567	179.03	1	181.115525	181.058645
200	2019-04-13	59567	179.36	1	181.103727	181.754481
202	2019-04-16	59567	178.61	2	182.384910	182.561780
203	2019-04-18	59567	177.77	1	182.026733	181.310542
204	2019-04-21	59567	177.77	1	182.556867	182.609471

# Autoregressive moving average(ARMA)
ARMA = ARIMA(train_df.price_per_sku.values, order=(1,0,1))
ARMA_model = ARMA.fit()

ARMA_model.predict(start =len(train_df), end= len(train_df) + len(test_df)-1)

array([179.97184905, 179.9823484 , 179.99237896, 180.00196165,
       180.01111649, 180.01986255, 180.02821811, 180.0362006 ,
       180.04382667, 180.05111223, 180.0580725 ])

test_df['ARMA_model_prediction'] = ARMA_model.predict(start =len(train_df), end= len(train_df) + len(test_df)-1)

# Autoregressive Integrated Moving Average
ARIMA_ = ARIMA(train_df.price_per_sku.values, order=(20,1,20))
ARIMA_model = ARIMA_.fit()

ARIMA_model.predict(start =len(train_df), end= len(train_df) + len(test_df)-1)

array([180.83419547, 180.48032031, 180.31605765, 181.11192182,
       178.6975036 , 179.13991888, 177.75230121, 177.77676229,
       179.94879324, 178.96750661, 180.37365968])

test_df['ARIMA_model_prediction'] = ARIMA_model.predict(start =len(train_df), end= len(train_df) + len(test_df)-1)

test_df

	dates	SKU	price_per_sku	num_purchases	AR_model_prediction	MA_model_prediction	ARMA_model_prediction	ARIMA_model_prediction
192	2019-04-02	59567	181.86	2	180.448574	181.141197	179.971849	180.834195
194	2019-04-06	59567	181.83	2	180.444365	180.971535	179.982348	180.480320
195	2019-04-08	59567	181.83	1	180.879267	181.784794	179.992379	180.316058
196	2019-04-09	59567	181.67	1	181.700718	182.459390	180.001962	181.111922
197	2019-04-10	59567	180.08	2	180.708146	181.104043	180.011116	178.697504
198	2019-04-11	59567	179.97	2	181.089813	181.896098	180.019863	179.139919
199	2019-04-12	59567	179.03	1	181.115525	181.058645	180.028218	177.752301
200	2019-04-13	59567	179.36	1	181.103727	181.754481	180.036201	177.776762
202	2019-04-16	59567	178.61	2	182.384910	182.561780	180.043827	179.948793
203	2019-04-18	59567	177.77	1	182.026733	181.310542	180.051112	178.967507
204	2019-04-21	59567	177.77	1	182.556867	182.609471	180.058073	180.373660

# Seasonal Autoregressive Integrated Moving-Average 
SARIMA = SARIMAX(train_df.price_per_sku.values, order=(20,1,20), seasonal= (0,0,0,0))
SARIMA_model = SARIMA.fit()

SARIMA_model.predict(start =len(train_df), end= len(train_df) + len(test_df)-1)

array([180.83419547, 180.48032031, 180.31605765, 181.11192182,
       178.6975036 , 179.13991888, 177.75230121, 177.77676229,
       179.94879324, 178.96750661, 180.37365968])

test_df['SARIMA_model_prediction'] = SARIMA_model.predict(start =len(train_df), end= len(train_df) + len(test_df)-1)

test_df

	dates	SKU	price_per_sku	num_purchases	AR_model_prediction	MA_model_prediction	ARMA_model_prediction	ARIMA_model_prediction	SARIMA_model_prediction
192	2019-04-02	59567	181.86	2	180.448574	181.141197	179.971849	180.834195	180.834195
194	2019-04-06	59567	181.83	2	180.444365	180.971535	179.982348	180.480320	180.480320
195	2019-04-08	59567	181.83	1	180.879267	181.784794	179.992379	180.316058	180.316058
196	2019-04-09	59567	181.67	1	181.700718	182.459390	180.001962	181.111922	181.111922
197	2019-04-10	59567	180.08	2	180.708146	181.104043	180.011116	178.697504	178.697504
198	2019-04-11	59567	179.97	2	181.089813	181.896098	180.019863	179.139919	179.139919
199	2019-04-12	59567	179.03	1	181.115525	181.058645	180.028218	177.752301	177.752301
200	2019-04-13	59567	179.36	1	181.103727	181.754481	180.036201	177.776762	177.776762
202	2019-04-16	59567	178.61	2	182.384910	182.561780	180.043827	179.948793	179.948793
203	2019-04-18	59567	177.77	1	182.026733	181.310542	180.051112	178.967507	178.967507
204	2019-04-21	59567	177.77	1	182.556867	182.609471	180.058073	180.373660	180.373660

# Simple Exponential Smoothing
SES = SimpleExpSmoothing(train_df.price_per_sku.values)
SES_model = SES.fit()

SES_model.predict(start =len(train_df), end= len(train_df) + len(test_df)-1)

array([179.96, 179.96, 179.96, 179.96, 179.96, 179.96, 179.96, 179.96,
       179.96, 179.96, 179.96])

test_df['SES_model_prediction'] = SES_model.predict(start =len(train_df), end= len(train_df) + len(test_df)-1)

test_df

	dates	SKU	price_per_sku	num_purchases	AR_model_prediction	MA_model_prediction	ARMA_model_prediction	ARIMA_model_prediction	SARIMA_model_prediction	SES_model_prediction
192	2019-04-02	59567	181.86	2	180.448574	181.141197	179.971849	180.834195	180.834195	179.96
194	2019-04-06	59567	181.83	2	180.444365	180.971535	179.982348	180.480320	180.480320	179.96
195	2019-04-08	59567	181.83	1	180.879267	181.784794	179.992379	180.316058	180.316058	179.96
196	2019-04-09	59567	181.67	1	181.700718	182.459390	180.001962	181.111922	181.111922	179.96
197	2019-04-10	59567	180.08	2	180.708146	181.104043	180.011116	178.697504	178.697504	179.96
198	2019-04-11	59567	179.97	2	181.089813	181.896098	180.019863	179.139919	179.139919	179.96
199	2019-04-12	59567	179.03	1	181.115525	181.058645	180.028218	177.752301	177.752301	179.96
200	2019-04-13	59567	179.36	1	181.103727	181.754481	180.036201	177.776762	177.776762	179.96
202	2019-04-16	59567	178.61	2	182.384910	182.561780	180.043827	179.948793	179.948793	179.96
203	2019-04-18	59567	177.77	1	182.026733	181.310542	180.051112	178.967507	178.967507	179.96
204	2019-04-21	59567	177.77	1	182.556867	182.609471	180.058073	180.373660	180.373660	179.96

# VAR
# VARMA
# VARMAX

import torch
from torch import nn
from torch.autograd import Variable as V
from torch import optim
import torchvision as tv
import torchvision.transforms as tf

class LSTM(nn.Module):
    def __init__(self, input_size=1, hidden_layer_size=100, output_size=1):
        super().__init__()
        self.hidden_layer_size = hidden_layer_size

        self.lstm = nn.LSTM(input_size, hidden_layer_size)

        self.linear = nn.Linear(hidden_layer_size, output_size)

        self.hidden_cell = (torch.zeros(1,1,self.hidden_layer_size),
                            torch.zeros(1,1,self.hidden_layer_size))

    def forward(self, input_seq):
        lstm_out, self.hidden_cell = self.lstm(input_seq.view(len(input_seq) ,1, -1), self.hidden_cell)
        predictions = self.linear(lstm_out.view(len(input_seq), -1))
        return predictions[-1]

model = LSTM()
loss_function = nn.MSELoss()
optimizer = torch.optim.Adam(model.parameters(), lr=0.001)

from sklearn.preprocessing import MinMaxScaler

scaler = MinMaxScaler(feature_range=(-1, 1))
train_data_normalized = scaler.fit_transform(train_df.price_per_sku.values .reshape(-1, 1))

train_data_normalized = torch.FloatTensor(train_data_normalized).view(-1)

train_data_normalized

tensor([-1.0000, -0.8011, -0.6172, -0.6172, -0.4325, -0.4238, -0.4238, -0.4238,
        -0.5927, -0.5927, -0.5927, -0.7680, -0.7624, -0.9116, -0.6093, -0.6030,
        -0.6748, -0.6748, -0.5051, -0.6946, -0.5998, -0.5998, -0.5762, -0.5841,
        -0.5841, -0.4736, -0.4736, -0.6527, -0.7285, -0.7285, -0.6938, -0.5525,
        -0.4230, -0.4230, -0.4878, -0.5122, -0.5122, -0.5122, -0.6275, -0.6275,
        -0.4002,  0.1902,  0.6156,  0.3236,  0.3197,  0.3197,  0.6819,  0.6819,
         0.6267,  0.6685,  0.6219,  0.6219,  0.6219,  1.0000,  0.7064,  0.7064,
         0.6093,  0.5122,  0.2313,  0.2313,  0.2313,  0.0892,  0.0892, -0.0426,
         0.2652,  0.2652,  0.3402,  0.2163,  0.1681,  0.1681,  0.0734,  0.0245,
        -0.0032,  0.0868,  0.0868, -0.0237,  0.0781,  0.0513,  0.0513,  0.0355,
         0.1263,  0.1263,  0.1263,  0.1294,  0.0371,  0.1168,  0.0545,  0.0545,
         0.0545,  0.2391,  0.2991,  0.2447,  0.2257,  0.2897,  0.3307,  0.3307,
         0.1807,  0.2202,  0.2202,  0.2644,  0.2912,  0.3717,  0.4286,  0.4286,
         0.6559,  0.7593,  0.9006,  0.9006,  0.9006,  0.3165,  0.3291,  0.3291,
         0.3291,  0.3891,  0.2952,  0.2249,  0.2005,  0.2005,  0.2076,  0.1279,
         0.1129,  0.2028,  0.0537,  0.0537,  0.0331,  0.0552,  0.1413,  0.0655,
         0.0576,  0.2470,  0.1823,  0.1721,  0.1721, -0.0347,  0.0442,  0.0868,
         0.0868,  0.1444, -0.2415, -0.3678, -0.3615, -0.3615, -0.1350, -0.1500,
        -0.1500])

def create_inout_sequences(input_data, tw=11):
    inout_seq = []
    L = len(input_data)
    for i in range(L-tw):
        train_seq = input_data[i:i+tw]
        train_label = input_data[i+tw:i+tw+1]
        inout_seq.append((train_seq ,train_label))
    return inout_seq

train_inout_seq = create_inout_sequences(train_data_normalized, tw=11)

train_inout_seq

[(tensor([-1.0000, -0.8011, -0.6172, -0.6172, -0.4325, -0.4238, -0.4238, -0.4238,
          -0.5927, -0.5927, -0.5927]),
  tensor([-0.7680])),
 (tensor([-0.8011, -0.6172, -0.6172, -0.4325, -0.4238, -0.4238, -0.4238, -0.5927,
          -0.5927, -0.5927, -0.7680]),
  tensor([-0.7624])),
 (tensor([-0.6172, -0.6172, -0.4325, -0.4238, -0.4238, -0.4238, -0.5927, -0.5927,
          -0.5927, -0.7680, -0.7624]),
  tensor([-0.9116])),
 (tensor([-0.6172, -0.4325, -0.4238, -0.4238, -0.4238, -0.5927, -0.5927, -0.5927,
          -0.7680, -0.7624, -0.9116]),
  tensor([-0.6093])),
 (tensor([-0.4325, -0.4238, -0.4238, -0.4238, -0.5927, -0.5927, -0.5927, -0.7680,
          -0.7624, -0.9116, -0.6093]),
  tensor([-0.6030])),
 (tensor([-0.4238, -0.4238, -0.4238, -0.5927, -0.5927, -0.5927, -0.7680, -0.7624,
          -0.9116, -0.6093, -0.6030]),
  tensor([-0.6748])),
 (tensor([-0.4238, -0.4238, -0.5927, -0.5927, -0.5927, -0.7680, -0.7624, -0.9116,
          -0.6093, -0.6030, -0.6748]),
  tensor([-0.6748])),
 (tensor([-0.4238, -0.5927, -0.5927, -0.5927, -0.7680, -0.7624, -0.9116, -0.6093,
          -0.6030, -0.6748, -0.6748]),
  tensor([-0.5051])),
 (tensor([-0.5927, -0.5927, -0.5927, -0.7680, -0.7624, -0.9116, -0.6093, -0.6030,
          -0.6748, -0.6748, -0.5051]),
  tensor([-0.6946])),
 (tensor([-0.5927, -0.5927, -0.7680, -0.7624, -0.9116, -0.6093, -0.6030, -0.6748,
          -0.6748, -0.5051, -0.6946]),
  tensor([-0.5998])),
 (tensor([-0.5927, -0.7680, -0.7624, -0.9116, -0.6093, -0.6030, -0.6748, -0.6748,
          -0.5051, -0.6946, -0.5998]),
  tensor([-0.5998])),
 (tensor([-0.7680, -0.7624, -0.9116, -0.6093, -0.6030, -0.6748, -0.6748, -0.5051,
          -0.6946, -0.5998, -0.5998]),
  tensor([-0.5762])),
 (tensor([-0.7624, -0.9116, -0.6093, -0.6030, -0.6748, -0.6748, -0.5051, -0.6946,
          -0.5998, -0.5998, -0.5762]),
  tensor([-0.5841])),
 (tensor([-0.9116, -0.6093, -0.6030, -0.6748, -0.6748, -0.5051, -0.6946, -0.5998,
          -0.5998, -0.5762, -0.5841]),
  tensor([-0.5841])),
 (tensor([-0.6093, -0.6030, -0.6748, -0.6748, -0.5051, -0.6946, -0.5998, -0.5998,
          -0.5762, -0.5841, -0.5841]),
  tensor([-0.4736])),
 (tensor([-0.6030, -0.6748, -0.6748, -0.5051, -0.6946, -0.5998, -0.5998, -0.5762,
          -0.5841, -0.5841, -0.4736]),
  tensor([-0.4736])),
 (tensor([-0.6748, -0.6748, -0.5051, -0.6946, -0.5998, -0.5998, -0.5762, -0.5841,
          -0.5841, -0.4736, -0.4736]),
  tensor([-0.6527])),
 (tensor([-0.6748, -0.5051, -0.6946, -0.5998, -0.5998, -0.5762, -0.5841, -0.5841,
          -0.4736, -0.4736, -0.6527]),
  tensor([-0.7285])),
 (tensor([-0.5051, -0.6946, -0.5998, -0.5998, -0.5762, -0.5841, -0.5841, -0.4736,
          -0.4736, -0.6527, -0.7285]),
  tensor([-0.7285])),
 (tensor([-0.6946, -0.5998, -0.5998, -0.5762, -0.5841, -0.5841, -0.4736, -0.4736,
          -0.6527, -0.7285, -0.7285]),
  tensor([-0.6938])),
 (tensor([-0.5998, -0.5998, -0.5762, -0.5841, -0.5841, -0.4736, -0.4736, -0.6527,
          -0.7285, -0.7285, -0.6938]),
  tensor([-0.5525])),
 (tensor([-0.5998, -0.5762, -0.5841, -0.5841, -0.4736, -0.4736, -0.6527, -0.7285,
          -0.7285, -0.6938, -0.5525]),
  tensor([-0.4230])),
 (tensor([-0.5762, -0.5841, -0.5841, -0.4736, -0.4736, -0.6527, -0.7285, -0.7285,
          -0.6938, -0.5525, -0.4230]),
  tensor([-0.4230])),
 (tensor([-0.5841, -0.5841, -0.4736, -0.4736, -0.6527, -0.7285, -0.7285, -0.6938,
          -0.5525, -0.4230, -0.4230]),
  tensor([-0.4878])),
 (tensor([-0.5841, -0.4736, -0.4736, -0.6527, -0.7285, -0.7285, -0.6938, -0.5525,
          -0.4230, -0.4230, -0.4878]),
  tensor([-0.5122])),
 (tensor([-0.4736, -0.4736, -0.6527, -0.7285, -0.7285, -0.6938, -0.5525, -0.4230,
          -0.4230, -0.4878, -0.5122]),
  tensor([-0.5122])),
 (tensor([-0.4736, -0.6527, -0.7285, -0.7285, -0.6938, -0.5525, -0.4230, -0.4230,
          -0.4878, -0.5122, -0.5122]),
  tensor([-0.5122])),
 (tensor([-0.6527, -0.7285, -0.7285, -0.6938, -0.5525, -0.4230, -0.4230, -0.4878,
          -0.5122, -0.5122, -0.5122]),
  tensor([-0.6275])),
 (tensor([-0.7285, -0.7285, -0.6938, -0.5525, -0.4230, -0.4230, -0.4878, -0.5122,
          -0.5122, -0.5122, -0.6275]),
  tensor([-0.6275])),
 (tensor([-0.7285, -0.6938, -0.5525, -0.4230, -0.4230, -0.4878, -0.5122, -0.5122,
          -0.5122, -0.6275, -0.6275]),
  tensor([-0.4002])),
 (tensor([-0.6938, -0.5525, -0.4230, -0.4230, -0.4878, -0.5122, -0.5122, -0.5122,
          -0.6275, -0.6275, -0.4002]),
  tensor([0.1902])),
 (tensor([-0.5525, -0.4230, -0.4230, -0.4878, -0.5122, -0.5122, -0.5122, -0.6275,
          -0.6275, -0.4002,  0.1902]),
  tensor([0.6156])),
 (tensor([-0.4230, -0.4230, -0.4878, -0.5122, -0.5122, -0.5122, -0.6275, -0.6275,
          -0.4002,  0.1902,  0.6156]),
  tensor([0.3236])),
 (tensor([-0.4230, -0.4878, -0.5122, -0.5122, -0.5122, -0.6275, -0.6275, -0.4002,
           0.1902,  0.6156,  0.3236]),
  tensor([0.3197])),
 (tensor([-0.4878, -0.5122, -0.5122, -0.5122, -0.6275, -0.6275, -0.4002,  0.1902,
           0.6156,  0.3236,  0.3197]),
  tensor([0.3197])),
 (tensor([-0.5122, -0.5122, -0.5122, -0.6275, -0.6275, -0.4002,  0.1902,  0.6156,
           0.3236,  0.3197,  0.3197]),
  tensor([0.6819])),
 (tensor([-0.5122, -0.5122, -0.6275, -0.6275, -0.4002,  0.1902,  0.6156,  0.3236,
           0.3197,  0.3197,  0.6819]),
  tensor([0.6819])),
 (tensor([-0.5122, -0.6275, -0.6275, -0.4002,  0.1902,  0.6156,  0.3236,  0.3197,
           0.3197,  0.6819,  0.6819]),
  tensor([0.6267])),
 (tensor([-0.6275, -0.6275, -0.4002,  0.1902,  0.6156,  0.3236,  0.3197,  0.3197,
           0.6819,  0.6819,  0.6267]),
  tensor([0.6685])),
 (tensor([-0.6275, -0.4002,  0.1902,  0.6156,  0.3236,  0.3197,  0.3197,  0.6819,
           0.6819,  0.6267,  0.6685]),
  tensor([0.6219])),
 (tensor([-0.4002,  0.1902,  0.6156,  0.3236,  0.3197,  0.3197,  0.6819,  0.6819,
           0.6267,  0.6685,  0.6219]),
  tensor([0.6219])),
 (tensor([0.1902, 0.6156, 0.3236, 0.3197, 0.3197, 0.6819, 0.6819, 0.6267, 0.6685,
          0.6219, 0.6219]),
  tensor([0.6219])),
 (tensor([0.6156, 0.3236, 0.3197, 0.3197, 0.6819, 0.6819, 0.6267, 0.6685, 0.6219,
          0.6219, 0.6219]),
  tensor([1.])),
 (tensor([0.3236, 0.3197, 0.3197, 0.6819, 0.6819, 0.6267, 0.6685, 0.6219, 0.6219,
          0.6219, 1.0000]),
  tensor([0.7064])),
 (tensor([0.3197, 0.3197, 0.6819, 0.6819, 0.6267, 0.6685, 0.6219, 0.6219, 0.6219,
          1.0000, 0.7064]),
  tensor([0.7064])),
 (tensor([0.3197, 0.6819, 0.6819, 0.6267, 0.6685, 0.6219, 0.6219, 0.6219, 1.0000,
          0.7064, 0.7064]),
  tensor([0.6093])),
 (tensor([0.6819, 0.6819, 0.6267, 0.6685, 0.6219, 0.6219, 0.6219, 1.0000, 0.7064,
          0.7064, 0.6093]),
  tensor([0.5122])),
 (tensor([0.6819, 0.6267, 0.6685, 0.6219, 0.6219, 0.6219, 1.0000, 0.7064, 0.7064,
          0.6093, 0.5122]),
  tensor([0.2313])),
 (tensor([0.6267, 0.6685, 0.6219, 0.6219, 0.6219, 1.0000, 0.7064, 0.7064, 0.6093,
          0.5122, 0.2313]),
  tensor([0.2313])),
 (tensor([0.6685, 0.6219, 0.6219, 0.6219, 1.0000, 0.7064, 0.7064, 0.6093, 0.5122,
          0.2313, 0.2313]),
  tensor([0.2313])),
 (tensor([0.6219, 0.6219, 0.6219, 1.0000, 0.7064, 0.7064, 0.6093, 0.5122, 0.2313,
          0.2313, 0.2313]),
  tensor([0.0892])),
 (tensor([0.6219, 0.6219, 1.0000, 0.7064, 0.7064, 0.6093, 0.5122, 0.2313, 0.2313,
          0.2313, 0.0892]),
  tensor([0.0892])),
 (tensor([0.6219, 1.0000, 0.7064, 0.7064, 0.6093, 0.5122, 0.2313, 0.2313, 0.2313,
          0.0892, 0.0892]),
  tensor([-0.0426])),
 (tensor([ 1.0000,  0.7064,  0.7064,  0.6093,  0.5122,  0.2313,  0.2313,  0.2313,
           0.0892,  0.0892, -0.0426]),
  tensor([0.2652])),
 (tensor([ 0.7064,  0.7064,  0.6093,  0.5122,  0.2313,  0.2313,  0.2313,  0.0892,
           0.0892, -0.0426,  0.2652]),
  tensor([0.2652])),
 (tensor([ 0.7064,  0.6093,  0.5122,  0.2313,  0.2313,  0.2313,  0.0892,  0.0892,
          -0.0426,  0.2652,  0.2652]),
  tensor([0.3402])),
 (tensor([ 0.6093,  0.5122,  0.2313,  0.2313,  0.2313,  0.0892,  0.0892, -0.0426,
           0.2652,  0.2652,  0.3402]),
  tensor([0.2163])),
 (tensor([ 0.5122,  0.2313,  0.2313,  0.2313,  0.0892,  0.0892, -0.0426,  0.2652,
           0.2652,  0.3402,  0.2163]),
  tensor([0.1681])),
 (tensor([ 0.2313,  0.2313,  0.2313,  0.0892,  0.0892, -0.0426,  0.2652,  0.2652,
           0.3402,  0.2163,  0.1681]),
  tensor([0.1681])),
 (tensor([ 0.2313,  0.2313,  0.0892,  0.0892, -0.0426,  0.2652,  0.2652,  0.3402,
           0.2163,  0.1681,  0.1681]),
  tensor([0.0734])),
 (tensor([ 0.2313,  0.0892,  0.0892, -0.0426,  0.2652,  0.2652,  0.3402,  0.2163,
           0.1681,  0.1681,  0.0734]),
  tensor([0.0245])),
 (tensor([ 0.0892,  0.0892, -0.0426,  0.2652,  0.2652,  0.3402,  0.2163,  0.1681,
           0.1681,  0.0734,  0.0245]),
  tensor([-0.0032])),
 (tensor([ 0.0892, -0.0426,  0.2652,  0.2652,  0.3402,  0.2163,  0.1681,  0.1681,
           0.0734,  0.0245, -0.0032]),
  tensor([0.0868])),
 (tensor([-0.0426,  0.2652,  0.2652,  0.3402,  0.2163,  0.1681,  0.1681,  0.0734,
           0.0245, -0.0032,  0.0868]),
  tensor([0.0868])),
 (tensor([ 0.2652,  0.2652,  0.3402,  0.2163,  0.1681,  0.1681,  0.0734,  0.0245,
          -0.0032,  0.0868,  0.0868]),
  tensor([-0.0237])),
 (tensor([ 0.2652,  0.3402,  0.2163,  0.1681,  0.1681,  0.0734,  0.0245, -0.0032,
           0.0868,  0.0868, -0.0237]),
  tensor([0.0781])),
 (tensor([ 0.3402,  0.2163,  0.1681,  0.1681,  0.0734,  0.0245, -0.0032,  0.0868,
           0.0868, -0.0237,  0.0781]),
  tensor([0.0513])),
 (tensor([ 0.2163,  0.1681,  0.1681,  0.0734,  0.0245, -0.0032,  0.0868,  0.0868,
          -0.0237,  0.0781,  0.0513]),
  tensor([0.0513])),
 (tensor([ 0.1681,  0.1681,  0.0734,  0.0245, -0.0032,  0.0868,  0.0868, -0.0237,
           0.0781,  0.0513,  0.0513]),
  tensor([0.0355])),
 (tensor([ 0.1681,  0.0734,  0.0245, -0.0032,  0.0868,  0.0868, -0.0237,  0.0781,
           0.0513,  0.0513,  0.0355]),
  tensor([0.1263])),
 (tensor([ 0.0734,  0.0245, -0.0032,  0.0868,  0.0868, -0.0237,  0.0781,  0.0513,
           0.0513,  0.0355,  0.1263]),
  tensor([0.1263])),
 (tensor([ 0.0245, -0.0032,  0.0868,  0.0868, -0.0237,  0.0781,  0.0513,  0.0513,
           0.0355,  0.1263,  0.1263]),
  tensor([0.1263])),
 (tensor([-0.0032,  0.0868,  0.0868, -0.0237,  0.0781,  0.0513,  0.0513,  0.0355,
           0.1263,  0.1263,  0.1263]),
  tensor([0.1294])),
 (tensor([ 0.0868,  0.0868, -0.0237,  0.0781,  0.0513,  0.0513,  0.0355,  0.1263,
           0.1263,  0.1263,  0.1294]),
  tensor([0.0371])),
 (tensor([ 0.0868, -0.0237,  0.0781,  0.0513,  0.0513,  0.0355,  0.1263,  0.1263,
           0.1263,  0.1294,  0.0371]),
  tensor([0.1168])),
 (tensor([-0.0237,  0.0781,  0.0513,  0.0513,  0.0355,  0.1263,  0.1263,  0.1263,
           0.1294,  0.0371,  0.1168]),
  tensor([0.0545])),
 (tensor([0.0781, 0.0513, 0.0513, 0.0355, 0.1263, 0.1263, 0.1263, 0.1294, 0.0371,
          0.1168, 0.0545]),
  tensor([0.0545])),
 (tensor([0.0513, 0.0513, 0.0355, 0.1263, 0.1263, 0.1263, 0.1294, 0.0371, 0.1168,
          0.0545, 0.0545]),
  tensor([0.0545])),
 (tensor([0.0513, 0.0355, 0.1263, 0.1263, 0.1263, 0.1294, 0.0371, 0.1168, 0.0545,
          0.0545, 0.0545]),
  tensor([0.2391])),
 (tensor([0.0355, 0.1263, 0.1263, 0.1263, 0.1294, 0.0371, 0.1168, 0.0545, 0.0545,
          0.0545, 0.2391]),
  tensor([0.2991])),
 (tensor([0.1263, 0.1263, 0.1263, 0.1294, 0.0371, 0.1168, 0.0545, 0.0545, 0.0545,
          0.2391, 0.2991]),
  tensor([0.2447])),
 (tensor([0.1263, 0.1263, 0.1294, 0.0371, 0.1168, 0.0545, 0.0545, 0.0545, 0.2391,
          0.2991, 0.2447]),
  tensor([0.2257])),
 (tensor([0.1263, 0.1294, 0.0371, 0.1168, 0.0545, 0.0545, 0.0545, 0.2391, 0.2991,
          0.2447, 0.2257]),
  tensor([0.2897])),
 (tensor([0.1294, 0.0371, 0.1168, 0.0545, 0.0545, 0.0545, 0.2391, 0.2991, 0.2447,
          0.2257, 0.2897]),
  tensor([0.3307])),
 (tensor([0.0371, 0.1168, 0.0545, 0.0545, 0.0545, 0.2391, 0.2991, 0.2447, 0.2257,
          0.2897, 0.3307]),
  tensor([0.3307])),
 (tensor([0.1168, 0.0545, 0.0545, 0.0545, 0.2391, 0.2991, 0.2447, 0.2257, 0.2897,
          0.3307, 0.3307]),
  tensor([0.1807])),
 (tensor([0.0545, 0.0545, 0.0545, 0.2391, 0.2991, 0.2447, 0.2257, 0.2897, 0.3307,
          0.3307, 0.1807]),
  tensor([0.2202])),
 (tensor([0.0545, 0.0545, 0.2391, 0.2991, 0.2447, 0.2257, 0.2897, 0.3307, 0.3307,
          0.1807, 0.2202]),
  tensor([0.2202])),
 (tensor([0.0545, 0.2391, 0.2991, 0.2447, 0.2257, 0.2897, 0.3307, 0.3307, 0.1807,
          0.2202, 0.2202]),
  tensor([0.2644])),
 (tensor([0.2391, 0.2991, 0.2447, 0.2257, 0.2897, 0.3307, 0.3307, 0.1807, 0.2202,
          0.2202, 0.2644]),
  tensor([0.2912])),
 (tensor([0.2991, 0.2447, 0.2257, 0.2897, 0.3307, 0.3307, 0.1807, 0.2202, 0.2202,
          0.2644, 0.2912]),
  tensor([0.3717])),
 (tensor([0.2447, 0.2257, 0.2897, 0.3307, 0.3307, 0.1807, 0.2202, 0.2202, 0.2644,
          0.2912, 0.3717]),
  tensor([0.4286])),
 (tensor([0.2257, 0.2897, 0.3307, 0.3307, 0.1807, 0.2202, 0.2202, 0.2644, 0.2912,
          0.3717, 0.4286]),
  tensor([0.4286])),
 (tensor([0.2897, 0.3307, 0.3307, 0.1807, 0.2202, 0.2202, 0.2644, 0.2912, 0.3717,
          0.4286, 0.4286]),
  tensor([0.6559])),
 (tensor([0.3307, 0.3307, 0.1807, 0.2202, 0.2202, 0.2644, 0.2912, 0.3717, 0.4286,
          0.4286, 0.6559]),
  tensor([0.7593])),
 (tensor([0.3307, 0.1807, 0.2202, 0.2202, 0.2644, 0.2912, 0.3717, 0.4286, 0.4286,
          0.6559, 0.7593]),
  tensor([0.9006])),
 (tensor([0.1807, 0.2202, 0.2202, 0.2644, 0.2912, 0.3717, 0.4286, 0.4286, 0.6559,
          0.7593, 0.9006]),
  tensor([0.9006])),
 (tensor([0.2202, 0.2202, 0.2644, 0.2912, 0.3717, 0.4286, 0.4286, 0.6559, 0.7593,
          0.9006, 0.9006]),
  tensor([0.9006])),
 (tensor([0.2202, 0.2644, 0.2912, 0.3717, 0.4286, 0.4286, 0.6559, 0.7593, 0.9006,
          0.9006, 0.9006]),
  tensor([0.3165])),
 (tensor([0.2644, 0.2912, 0.3717, 0.4286, 0.4286, 0.6559, 0.7593, 0.9006, 0.9006,
          0.9006, 0.3165]),
  tensor([0.3291])),
 (tensor([0.2912, 0.3717, 0.4286, 0.4286, 0.6559, 0.7593, 0.9006, 0.9006, 0.9006,
          0.3165, 0.3291]),
  tensor([0.3291])),
 (tensor([0.3717, 0.4286, 0.4286, 0.6559, 0.7593, 0.9006, 0.9006, 0.9006, 0.3165,
          0.3291, 0.3291]),
  tensor([0.3291])),
 (tensor([0.4286, 0.4286, 0.6559, 0.7593, 0.9006, 0.9006, 0.9006, 0.3165, 0.3291,
          0.3291, 0.3291]),
  tensor([0.3891])),
 (tensor([0.4286, 0.6559, 0.7593, 0.9006, 0.9006, 0.9006, 0.3165, 0.3291, 0.3291,
          0.3291, 0.3891]),
  tensor([0.2952])),
 (tensor([0.6559, 0.7593, 0.9006, 0.9006, 0.9006, 0.3165, 0.3291, 0.3291, 0.3291,
          0.3891, 0.2952]),
  tensor([0.2249])),
 (tensor([0.7593, 0.9006, 0.9006, 0.9006, 0.3165, 0.3291, 0.3291, 0.3291, 0.3891,
          0.2952, 0.2249]),
  tensor([0.2005])),
 (tensor([0.9006, 0.9006, 0.9006, 0.3165, 0.3291, 0.3291, 0.3291, 0.3891, 0.2952,
          0.2249, 0.2005]),
  tensor([0.2005])),
 (tensor([0.9006, 0.9006, 0.3165, 0.3291, 0.3291, 0.3291, 0.3891, 0.2952, 0.2249,
          0.2005, 0.2005]),
  tensor([0.2076])),
 (tensor([0.9006, 0.3165, 0.3291, 0.3291, 0.3291, 0.3891, 0.2952, 0.2249, 0.2005,
          0.2005, 0.2076]),
  tensor([0.1279])),
 (tensor([0.3165, 0.3291, 0.3291, 0.3291, 0.3891, 0.2952, 0.2249, 0.2005, 0.2005,
          0.2076, 0.1279]),
  tensor([0.1129])),
 (tensor([0.3291, 0.3291, 0.3291, 0.3891, 0.2952, 0.2249, 0.2005, 0.2005, 0.2076,
          0.1279, 0.1129]),
  tensor([0.2028])),
 (tensor([0.3291, 0.3291, 0.3891, 0.2952, 0.2249, 0.2005, 0.2005, 0.2076, 0.1279,
          0.1129, 0.2028]),
  tensor([0.0537])),
 (tensor([0.3291, 0.3891, 0.2952, 0.2249, 0.2005, 0.2005, 0.2076, 0.1279, 0.1129,
          0.2028, 0.0537]),
  tensor([0.0537])),
 (tensor([0.3891, 0.2952, 0.2249, 0.2005, 0.2005, 0.2076, 0.1279, 0.1129, 0.2028,
          0.0537, 0.0537]),
  tensor([0.0331])),
 (tensor([0.2952, 0.2249, 0.2005, 0.2005, 0.2076, 0.1279, 0.1129, 0.2028, 0.0537,
          0.0537, 0.0331]),
  tensor([0.0552])),
 (tensor([0.2249, 0.2005, 0.2005, 0.2076, 0.1279, 0.1129, 0.2028, 0.0537, 0.0537,
          0.0331, 0.0552]),
  tensor([0.1413])),
 (tensor([0.2005, 0.2005, 0.2076, 0.1279, 0.1129, 0.2028, 0.0537, 0.0537, 0.0331,
          0.0552, 0.1413]),
  tensor([0.0655])),
 (tensor([0.2005, 0.2076, 0.1279, 0.1129, 0.2028, 0.0537, 0.0537, 0.0331, 0.0552,
          0.1413, 0.0655]),
  tensor([0.0576])),
 (tensor([0.2076, 0.1279, 0.1129, 0.2028, 0.0537, 0.0537, 0.0331, 0.0552, 0.1413,
          0.0655, 0.0576]),
  tensor([0.2470])),
 (tensor([0.1279, 0.1129, 0.2028, 0.0537, 0.0537, 0.0331, 0.0552, 0.1413, 0.0655,
          0.0576, 0.2470]),
  tensor([0.1823])),
 (tensor([0.1129, 0.2028, 0.0537, 0.0537, 0.0331, 0.0552, 0.1413, 0.0655, 0.0576,
          0.2470, 0.1823]),
  tensor([0.1721])),
 (tensor([0.2028, 0.0537, 0.0537, 0.0331, 0.0552, 0.1413, 0.0655, 0.0576, 0.2470,
          0.1823, 0.1721]),
  tensor([0.1721])),
 (tensor([0.0537, 0.0537, 0.0331, 0.0552, 0.1413, 0.0655, 0.0576, 0.2470, 0.1823,
          0.1721, 0.1721]),
  tensor([-0.0347])),
 (tensor([ 0.0537,  0.0331,  0.0552,  0.1413,  0.0655,  0.0576,  0.2470,  0.1823,
           0.1721,  0.1721, -0.0347]),
  tensor([0.0442])),
 (tensor([ 0.0331,  0.0552,  0.1413,  0.0655,  0.0576,  0.2470,  0.1823,  0.1721,
           0.1721, -0.0347,  0.0442]),
  tensor([0.0868])),
 (tensor([ 0.0552,  0.1413,  0.0655,  0.0576,  0.2470,  0.1823,  0.1721,  0.1721,
          -0.0347,  0.0442,  0.0868]),
  tensor([0.0868])),
 (tensor([ 0.1413,  0.0655,  0.0576,  0.2470,  0.1823,  0.1721,  0.1721, -0.0347,
           0.0442,  0.0868,  0.0868]),
  tensor([0.1444])),
 (tensor([ 0.0655,  0.0576,  0.2470,  0.1823,  0.1721,  0.1721, -0.0347,  0.0442,
           0.0868,  0.0868,  0.1444]),
  tensor([-0.2415])),
 (tensor([ 0.0576,  0.2470,  0.1823,  0.1721,  0.1721, -0.0347,  0.0442,  0.0868,
           0.0868,  0.1444, -0.2415]),
  tensor([-0.3678])),
 (tensor([ 0.2470,  0.1823,  0.1721,  0.1721, -0.0347,  0.0442,  0.0868,  0.0868,
           0.1444, -0.2415, -0.3678]),
  tensor([-0.3615])),
 (tensor([ 0.1823,  0.1721,  0.1721, -0.0347,  0.0442,  0.0868,  0.0868,  0.1444,
          -0.2415, -0.3678, -0.3615]),
  tensor([-0.3615])),
 (tensor([ 0.1721,  0.1721, -0.0347,  0.0442,  0.0868,  0.0868,  0.1444, -0.2415,
          -0.3678, -0.3615, -0.3615]),
  tensor([-0.1350])),
 (tensor([ 0.1721, -0.0347,  0.0442,  0.0868,  0.0868,  0.1444, -0.2415, -0.3678,
          -0.3615, -0.3615, -0.1350]),
  tensor([-0.1500])),
 (tensor([-0.0347,  0.0442,  0.0868,  0.0868,  0.1444, -0.2415, -0.3678, -0.3615,
          -0.3615, -0.1350, -0.1500]),
  tensor([-0.1500]))]

epochs = 80

for i in range(epochs):
    for seq, labels in train_inout_seq:
        optimizer.zero_grad()
        model.hidden_cell = (torch.zeros(1, 1, model.hidden_layer_size),
                        torch.zeros(1, 1, model.hidden_layer_size))

        y_pred = model(seq)

        single_loss = loss_function(y_pred, labels)
        single_loss.backward()
        optimizer.step()

    if i%10 == 1:
        print(f'epoch: {i:3} loss: {single_loss.item():10.8f}')

print(f'epoch: {i:3} loss: {single_loss.item():10.10f}')

epoch:   1 loss: 0.02080892
epoch:  11 loss: 0.00310560
epoch:  21 loss: 0.00071738
epoch:  31 loss: 0.00000512
epoch:  41 loss: 0.00000000
epoch:  51 loss: 0.00047642
epoch:  61 loss: 0.00004787
epoch:  71 loss: 0.01555970
epoch:  79 loss: 0.0173519235

test_inputs = train_data_normalized[-11:].tolist()

model.eval()

for i in range(11):
    seq = torch.FloatTensor(test_inputs[-11:])
    with torch.no_grad():
        model.hidden = (torch.zeros(1, 1, model.hidden_layer_size),
                        torch.zeros(1, 1, model.hidden_layer_size))
        test_inputs.append(model(seq).item())

test_inputs[11:]

[-0.17209258675575256,
 -0.16259273886680603,
 -0.12296204268932343,
 -0.04300084710121155,
 0.04384655877947807,
 0.125918447971344,
 0.2013656347990036,
 0.27468109130859375,
 0.3663773238658905,
 0.45008495450019836,
 0.5124920010566711]

https://machinelearningmastery.com/grid-search-arima-hyperparameters-with-python/

https://machinelearningmastery.com/time-series-forecasting-methods-in-python-cheat-sheet/

План прогноза спроса

Расчет эластичности не примере

cnt= [i for i in range(0,110,10)]
price = sorted([i for i in range(1,11)], reverse=True)

df = pd.DataFrame(zip(cnt,price), columns=['cnt', 'price'])

df

	cnt	price
0	0	10
1	10	9
2	20	8
3	30	7
4	40	6
5	50	5
6	60	4
7	70	3
8	80	2
9	90	1

$E_{pd} = \frac{\Delta Q}{\Delta P}*\frac{P}{Q} = \frac{(P_1+P_2)\Delta Q}{(Q_1+Q_2)\Delta P}$

df['prev_price'] = df.price.shift()
df['prev_cnt'] = df.cnt.shift()

df

	cnt	price	prev_price	prev_cnt
0	0	10	NaN	NaN
1	10	9	10.0	0.0
2	20	8	9.0	10.0
3	30	7	8.0	20.0
4	40	6	7.0	30.0
5	50	5	6.0	40.0
6	60	4	5.0	50.0
7	70	3	4.0	60.0
8	80	2	3.0	70.0
9	90	1	2.0	80.0

df['elasticity_coef'] = None
df.loc[~df.prev_price.isnull(), 'elasticity_coef'] = np.abs(((df.prev_price+df.price)*(df.cnt-df.prev_cnt))
                                                            /((df.cnt+df.prev_cnt)*(df.price-df.prev_price)))

df

	cnt	price	prev_price	prev_cnt	elasticity_coef
0	0	10	NaN	NaN	None
1	10	9	10.0	0.0	19.0
2	20	8	9.0	10.0	5.666667
3	30	7	8.0	20.0	3.0
4	40	6	7.0	30.0	1.857143
5	50	5	6.0	40.0	1.222222
6	60	4	5.0	50.0	0.818182
7	70	3	4.0	60.0	0.538462
8	80	2	3.0	70.0	0.333333
9	90	1	2.0	80.0	0.176471

plt.figure(figsize=(12,8))
sns.lineplot(x =df['cnt'], y=df['price'])

<matplotlib.axes._subplots.AxesSubplot at 0x7fbcce0a7110>

План разработки приложения для ценообразования

1) Выбрать товары для отдельного ценообразования

2) Для данных товаров построить модели предсказания спроса в зависимости от цены

3) Согласно выбранной стратегии компании выбрать, что именно мы будем оптимизировать

4) Корректировка цен относительно ограничений и цен конкурентов

5) Получение обратной связи по изменение спроса в зависимости от цены и возврат к пункту 2