# use machine learning to predict a next day of visit for customer

I have a problem a need your suggestion , I am working in a retail data , and want to predict the behavior of the customer , the data contains information about the customer who visits the shopping center, it only contains the customer unique identifier and the customer visits for 143 weeks so for each record I have visitor Id and visits , the visits contains numbers such as 1 5 30 ..etc , 1 is the first day of the first week which is Monday then 5 means Friday , then 30 means Tuesday for the fourth week and etc , what I am trying to do is to predict when the first day of the visit for the next 144th week can you please help me in this

This is the data format

+------------+-------------------------------------------------------------------------------------------------------------------+
| visitor_id |                                                       visits                                                      |
+------------+-------------------------------------------------------------------------------------------------------------------+
|          1 |             30 84 126 135 137 179 242 342 426 456 460 462 483 594 600 604 704 723 744 787 804 886 924 928 946 954 |
|          2 |  24 53 75 134 158 192 194 211 213 238 251 305 404 418 458 476 493 571 619 731 739 759 761 847 883 943 962 981 983 |
+------------+-------------------------------------------------------------------------------------------------------------------+

• Have a look at this historic answer datascience.stackexchange.com/q/17099/35644, Have a look at this kaggle comp kaggle.com/c/demand-forecasting-kernels-only/kernels, Sep 10 '18 at 12:04
• What are your features that you can use to predict the first day of the visit for the next 144th week? These visits to me are rather your labels. And your problem looks like a multi-label (multi-output) classification. Meaning you can create a dummy columns out visits and try to predict which occurs? But I am confused about your visits. Should not you only have max 364 days for one year? Why you have a lot more than that? Please edit your question with more clear data so people may be able to help. Oct 17 '18 at 4:51

You can approach that problem as time-series prediction problem. Below is one of articles or tutorials about time-series forecasting/prediction.

https://machinelearningmastery.com/time-series-forecasting/

Objective: Predict if a visitor_id will visit the retail store(on 144th day)

Data: Visitor ID and Visited day Number

Approach:

1. Find unique number of Visitors and group them by number of visits

2. Select by group

3. Transform Data into individual series as follows: example

For Visitor_ID: X1
Day No      Visited(Yes or No)
1           Yes
2           Yes
3           No
4           No
.           .
.           .
.           .
.           .
143         No

For Visitor_ID: X2
Day No      Visited(Yes or No)
1           No
2           Yes
3           Yes
4           No
.           .
.           .
.           .
.           .
143         Yes

4. Look for suitable algorithm for the data for prediction.

Hope this helps.