Customer Churn Prediction and Retention through Personalised Recommendation System in a SuperMarket
Name
Prashanth Parthiban
Abstract
Customer churn has been a key area of revenue loss for retailers specifically
when it concerns an offline market. As customers are not bound by any contract,
it is often the case that they are lost to the whims of discounts and incentives
offered by competitors.
In order to curtail this situation we suggest a framework wherein customers who
are going to churn in 3-6 months are identified well in advance with supervised
machine learning approach. Once churners are identified we train a recommendation
system based on their transactional history to suggest products and therefore
prevent churners from churning.
In this paper, a novel algorithmic framework is suggested to overcome the
churn issue with the help of recommendation system. The most effective way to
identify a churner is based on RFM (Recency, Frequency and Money) features.
The models are built on various features about the customer and their shopping
habits in the past. Identifying the right algorithm which serves the purpose is of
utmost importance and for that we apply and test the performance of quite a few
algorithms namely Random Forest, K-Nearest Neighbors, Decision Tree, Gradient
Boosting Method.
Recommender Model applied are User Based Collaborative Filtering and Item
Based Recommmender System. Experiments are performed on real market data
to prove the effectiveness of proposed framework. Thus with the help of churn and recommender model, churners are identified and retained.
when it concerns an offline market. As customers are not bound by any contract,
it is often the case that they are lost to the whims of discounts and incentives
offered by competitors.
In order to curtail this situation we suggest a framework wherein customers who
are going to churn in 3-6 months are identified well in advance with supervised
machine learning approach. Once churners are identified we train a recommendation
system based on their transactional history to suggest products and therefore
prevent churners from churning.
In this paper, a novel algorithmic framework is suggested to overcome the
churn issue with the help of recommendation system. The most effective way to
identify a churner is based on RFM (Recency, Frequency and Money) features.
The models are built on various features about the customer and their shopping
habits in the past. Identifying the right algorithm which serves the purpose is of
utmost importance and for that we apply and test the performance of quite a few
algorithms namely Random Forest, K-Nearest Neighbors, Decision Tree, Gradient
Boosting Method.
Recommender Model applied are User Based Collaborative Filtering and Item
Based Recommmender System. Experiments are performed on real market data
to prove the effectiveness of proposed framework. Thus with the help of churn and recommender model, churners are identified and retained.
Graduation Thesis language
English
Graduation Thesis type
Master - Software Engineering
Supervisor(s)
Anna Leontjeva
Defence year
2016