Predictive Model

Dear all,

I would like to ask you how can I build a predictive model that predict the sales of the next year with respect to the data provided of 2017,2018 and 2019. I would like also to filter out the customers who have the high probability to order again, for this I have the phones number of the customers that are ordered in 2017, 2018 and 2019. I have attached the sample excel sheet of the data that I have and according to this data I want to build the predictive model. So, any help in that please?

Note that: The data that I want to predict the sales of it it's from C column to N column.

I delete the phones number of the customer because of the privacy.

Thanks in advance for your kind support.

Sample.xlsx

Workflow

Predictive Analysis

Udacity

Accepted answers

RolandSchubert

Hi @Feras95p ,

I've changed the customer identifier to "mobile", if you use data containing mobile number it should work (hopefully ...). I also added the replace NULL by value 0 to solve the problem with NULL values

Best,

Roland

CustomerReturn.yxmd

RolandSchubert

I select the customers with visit count > 1 (for all other customers an average makes no sense).

Next step is sorting by mobile (= customer) and date. The Multi-Row Formula tool calculates the difference between subsequent rows using the formula DateTimeDiff([Date], [Row-1:Date], 'days'). Selecting [mobile phone] in the "Group by" option, ensures that the difference between visits of each customer is calculated separately.

As the first row has no predecessor, the difference is NULL for this row, so I filter them out by using a filter tool (added this in the new version)

To calculate the average, the Summarize tool is used - Group by mobile (= customer) and average for difference.

CustomerReturn.yxmd

All comments

echuong1

Take a look at the tool mastery page for the linear regression tool:

https://community.alteryx.com/t5/Alteryx-Designer-Knowledge-Base/Tool-Mastery-Linear-Regression/ta-p/35464

Other types of models can be found in the index:

https://community.alteryx.com/t5/Alteryx-Designer-Knowledge-Base/Tool-Mastery-Index/ta-p/84593

wdavis

Hi @Feras95p

There is also an excellent article here - https://community.alteryx.com/t5/Data-Science-Blog/Champagne-Analytics-A-Time-Series-Tutorial/ba-p/419776 which documents how you can use the Time Series tools in Alteryx.

Thanks

Will

RolandSchubert

Hi @Feras95p ,

I think you have to build two different model:

1) Predict sales based on history: In my opinion, you should aggregate the values for columns C-N on a daily level and use one of the Time Series tools to create a predictive model (ETS, ARIMA). You'll have to build one model for each column.

2) Predict probability for customers: You have to identify the most important variables for customers to return - e.g. frequency (number of visits) and recency (last visit), period between visits.Maybe it's possible to build a regession model based on these (probability for return of customer).

What do you think?

Best,

Roland

Feras95p

1- The values of columns C-N are already in daily level, but I don't know how I can build the predictive model so, if you can explain for me the steps for one model and I will do for the other it will be much appreciated.

2- I have already identified how many each customers return to order again but I don't know how I can estimate the avg time that they took to return, so I want to guide me on how I can do that.

afv2688

Hello @Feras95p ,

Check the ETS and ARIMA modles in the designer. There are many articles whithin the community that can help you to configure them:

https://help.alteryx.com/2018.2/ETS.htm

https://help.alteryx.com/current/ARIMA.htm

Edit:

You can also check these examples:

https://community.alteryx.com/t5/Data-Science-Blog/Expand-Your-Predictive-Palette-III-I-Sales-Forecast-with-Prophet/ba-p/504651

https://community.alteryx.com/t5/Data-Science-Blog/Champagne-Analytics-A-Time-Series-Tutorial/ba-p/419776

If this post helps, then please consider Accept it as the solution to help the other members find it more quickly.

Regards

Quick Links

This months top contributors

atcodedog05 19458

Qiu 15866

binu_acs 15708

MarqueeCrew 13708

apathetichell 13703