A FRAMEWORK FOR PREDICTING ONLINE BEHAVIOUR OF THE USERS USING CLICK STREAM

Customer Relationship Management systems have been used to allow businesses to attract new customers, develop a long - term relationship with them and improve the retention of customers for greater profitability. CRM systems use machine-learning models to evaluate personal and behavioural data of customers to give company a competitive edge by increasing the retention rate of customers. These models can predict customer's purchases, and their reasons. Predictions are used in the development of targeted marketing plans and service offerings. This paper focuses to develop a framework by applying machine learning techniques to predict the purchases by the customers on e-commerce platform. Through clickstream and additional customer data, frameworks for predicting customer behaviour can indeed be developed. Predicting potential consumer behaviour generates pertinent information for sales and marketing teams to strategically focus on different resources. Such information facilitates inventory planning at the warehouse and at the point of sale, as well as strategic decisions during production processes can be planned accordingly. Next, this research provides insight into the performance differences of the models on sequential clickstream and static customer data by performing a data analysis and training the models separately on the various datasets. Deep Belief network along with auto encoders and Adam optimizer is performed to validate the data and predict the likelihood of purchase and customer conversion probability.