Customer Repurchase accurate prediction

This is the major problem which is faced by many companies . You may think your shop or MNC is running well today ! But , Will it run well tommorow or next year ? After seeing the cases of Jet Airways , Reliance Telecom , Caffe Coffee Day etc tragedies . You must realize the importance of data miners.

I have built a model doing literature survey which can be improvised with future advancements. How do you think customer behaves and analyzes for buying a product? Let me assume myself as a customer and list down the points which excites me for shopping. # Sale offer# Warranty #new designs #need of an hour etc. If we go on listing the facts from customer lifestyle, it will just on. The model which I would like to present is the below one:

In this figure, I have differentiated them of offline and Online features which may likely lead to customer Repurchase. Allow me to present you with the steps of execution. After that, I will explain you in a detailed format.

🤷For those people who are wondering what is pilot test after going through this flowchart.

The survey method of collecting data will go by answering questionaire. Once you shop in big brands such as FOREVER 21 , WESTSIDE , LIFESTYLE etc while paying the amount. They eventually request you to fill the survey questions. We got our data through this mode.

🛑 Usage of ABC algorithm ( artificial bee colony theorem) — Works on three phases ( Employee bees, onlookers bees and scout bees) . Basically a cluster algorithm which uses Sum of squared error & centroids play a major rule.

We are using this ABC clustering & optimization algorithm over the features collected by customer lifestyle. Then we will use ML algorithms for accuracy, sensitivity etc.

As you are planning to try this model, I would like to help you with starting steps in coding. The above codes will start your journey happily.

🛑 Correlation Analysis

We are trying out the pairs which are dependent on each other such height-weight etc.

Best is to plot the data and check the correlation. As it will provide a brighter idea. It’s better to take one more step before applying ABC algorithm. This is called as Cronbach alpha. It is necessary because :

🛑 K- cross validation :

Obviously, we are dividing the data into testing and training. You must know, why we are going with it?

Now , it’s action time !

After applying ABC algorithm, we got this following results exclusively out of all the features.

This is the moment, we should go ahead with Decision tree, Random decision forest , neutral networks etc for analysing the best machine algorithm for customer Repurchase prediction. I would recommend people to go ahead with adaboost algorithm to support decision. Finally, you can plot graphs along evaluation parameters such as KS graph, RMSE etc.

🙋 Future advancement:

🛑 Building of emotional face app such as smile, angry detectors using Cascade classifier. And note down the customer response from current shopping.

🛑Using speed recognition by tensorflow and provide staffless assistance for cash payment.