Design And Implementation Of Scoring System Based On Intelligent Recommendation Algorithm

It has been brought to the classical e-Commerce system with a great impact, customers now have more possibility to get to know about the items and make them acquire the most important place in modern e-Commerce. Shopping trends or fashion should be discovered much quickly as for the e-Commerce system nowadays. This need promoted the broad development in the recommender system in recent decades. Its application is not only restricted in determining advertise strategies and promoting sales account with helping customers know the items better, recommender system is also utilized for improving sales strategies in a form of feedback, especially in some magnate company like Amazon. The management of the whole company is now based on the reflection of the recommender system running. The design of a recommender system crosses over many research area like statistics and probability, machine learning techniques and psychology theorems. Machine learning algorithms keep the system automatic, intelligent and the learning method can reduce the predictive error at some point, which are therefore broadly applied in implementing recommender system.This thesis first give some descriptions of the relative knowledge and background terms in designing recommender system and some recommender algorithms based on machine learning. It is followed with the implementation of a common recommender system framework with analysis of some algorithms in detail. The philosophy and implementation detail of every recommender algorithm is greatly delved. Considering the particular need of business company, a complete intelligent e-Commerce system with recommender is fully implemented for testing all provided algorithms. The system contains independent database server and recommender server based on the framework introduced above. The great performance of the system in a real-life dataset has proved its availability. The final recommender system fulfills the requirements and can be extended and deployed to other platforms with its design in module and high scalability.