Research And Application Of Sorting Recommend Algorithms With Multiple Factors

With the development of 020 platforms has been used, more people choose to travel in China orders a taxi under the platform software, and more people choose to become a platformer for users to take advantage of a convenient time to use platform orders. In this context, how to improve the getting orders rating will be the most appropriate for the current user's order pushed to showing in front of users as soon as possible, and how to allowing users to receive their most appropriate order is a big problem now. Since traditional sort order recommended way is to simply push the order, in the small user groups and relatively accurate sort order size can also recommend, but as more and more people are choosing to use the platform, the traditional sort order recommended way is not suitable for the current scale of large user groups and orders.According to the sort order featured in today's recommendations and personalized deficiencies, based on user behavior analysis of log data presented sort personalized recommendation algorithm based on user preferences. Based on the algorithm, paper for has user behavior data analysis and the user behavior analysis, from around with user of preferences for has user behavior features building, and combined Street features and orders features, will its into to user of behavior features in the building time, and street and orders of cross features as user of eventually behavior features, and effective practice in the should consider of route select problem and user behavior time sequence sex problem, And full consider has user in route and orders selected based on itself factors of consider by made of personalized behavior, according to user behavior of time sex factors proposed has based on user preference of part features dynamic update algorithm, based on this algorithm in user features in the joined has more of user personalized features, makes sort recommended more can meet user of itself needs, then paper in based on this algorithm building of user features based Shang, for based on user behavior of logic return model and support vector machine model of building, And through cross validation of method for algorithm parameter and model parameter of adjustment and validation, last on traditional of orders sort recommended and paper by created of based on machine learning model of orders sort recommended for has compared, experiment results compared surface, this experiment using of based on more factors of sort recommended algorithm for orders recommended more with accuracy, more can meet user in consider itself factors by reflected of personalized needs.