Customer behavior prediction
Μοντέλα πρόβλεψης συμπεριφοράς καταναλωτή
KeywordsDeep learning ; Machine learning ; Customer behavior ; Prediction ; Sequence to sequence ; Python ; TensorFlow2 ; Neural network ; LSTM ; Keras ; Churn ; Classification ; CX
As the science of Machine Learning evolves, it is a fact that finds the answers to many business needs. The huge amount of data, the fast way of collection and the fast processes of analysis are the main triggers to this. Day by day, a lot of companies already realize the power of data analytics. The subject of this dissertation is the behavioral prediction of a customer around a product/service, taking into consideration the sequence of activities of the customer on it. Briefly, we are trying to predict customer’s the next activity and churn alert flag through a sequence of activities using the python programming language. The results show that deep learning techniques can guide companies to new strategies and create targeting campaigns of communication or sale. We used many tools and methods to implement this thesis. First, we used R and RStudio to fetch the necessary first party data and to transform them more easily to the desired structure. The main work is done using python, and more specifically, we focused on Tensorflow2, Keras libraries and we used the LSTM technique. At the end, we mention the importance of visualization and we try to implement a web app to serve the results of the model with an easy and self-service way to the possible stakeholders. As of conclusion, we talk about the output of each experiment and mention some thoughts for future work.