A hybrid recommendation system that produces higher quality recommendations based on both content-based filtering and collaborative filtering approaches. At the same time, it is a solid and reliable project and documentation from beginning to end, coded with Python.
In this study, we propose and implement a novel hybrid recommender system. We give theoretical framework and related implementation in detail throughout this study. We conducted an experiment analyzing prediction quality of the proposal, as well. Our experimental results demonstrate that this hybrid scheme gives promising accuracy. In addition to accuracy, this proposal is efficient. Because both recommender approaches operate on item- based similarity computation, which enables item to item similarity matrix construction, offline. Prediction computation is needed when the system is running online. With the data update characteristics of web sites, offline similarity models should be updated periodically
Our software project about recommender systems has been published by Authorea Preprints. You can access from https://lnkd.in/d9S7N9CF.
Full implementation can be accessed from GitHub: https://lnkd.in/dqKpHzKF
