Recommender System
This is a project that aims at building a recommendation system to recommend new products out of the 25 available ones to a total of 46779 users. The available information we have includes:
- Users' characteristics such as country, gender, age...etc.
- Users' historical purchase records on a monthly basis during the time range from January 2015 to April 2016.
Based on the information we have and our understanding of a recommendation system, we decided to run 3 models to the dataset plus 1 model that takes into account all 3 models' results.
This is because we would like to know if the purchase behavior of the users is more affected by the users' characteristics, or by those who share similar purchase behaviors.
Apart from that, we also implemented a non-personalized recommendation model that, in any case, would be a good indication, especially under the situation when there's not much data for us to rely on.
Finally, we made predictions based on these 4 models and compared the MAP (Mean Average Precision) score among these 4 models to reach our conclusion of this project.
The steps to carry out this project at a higher level point of view are as follows:
- Data Exploration and Data Preprocessing
- Model Selection and Application
- Prediction and Model Evaluation