The Distributed Optimization Design And Implementation Of Restaurant Recommendation System Based On Hadoop

With the rapid development of Internet big data analysis platform,the user data analysis in all walks of life,especially those related to people's livelihood,has set off a wave of reform."Big data" analysis enables us to make effective use of our fragmented time,and the traditional lifestyle is undergoing disruptive changes.The core content of the current recommendation system research is how to effectively obtain the required information in the massive information accurately and recommend the required information to the intended population effectively,so as to make the information acquisition and search more efficient.Based on the habits of users in the restaurant consumption,this thesis conducts a research on the recommendation system for the two key attributes of users' rating of restaurants and the geographical coordinate information of restaurants.By analyzing the current research environment of the recommendation system at local and abroad and the classical algorithms used in the recommendation system,the item-based collaborative filtering algorithm and Geo Hash algorithm with the highest correlation with key attributes are selected for in-depth research and application.The similarity between the restaurants is calculated by the user's rating of restaurants,and the recommendation score is obtained by the similarity.Then,the real-time coordinate of the user and the coordinate of restaurants are rapidly matched by the Geo Hash algorithm,so as to recommend the list of restaurant which matched the requirement of the user.In standalone server operation environment,when dealing with large data as a result of the limitation of hardware cannot recommend results quickly,so this thesis presents a optimized plan by using distributed storage and operation framework based on Hadoop platform to improve collaborative filtering algorithm based on item and Geo Hash algorithm,and the optimized recommendation algorithm is verified by experiment through scheduling multiple servers in the cluster parallel computing.Comparing the running time under standalone server environment,the running time significantly lower under the environment of optimized plan,the running efficiency improved significantly.The purpose of this thesis,quickly recommend restaurants to target users,has been achieved.