Research On Bus Cloud Recommendation System Based On Machine Learning Algorithms

The rapid development of network and information technology has led to the period of information overload,and it is difficult for information consumers to get useful information efficiently from the complicated information.At the same time,information producers also expect that the information they produce can be pushed to the desired information consumers in a timely and accurate manner to attract their attention.Based on this,recommendation systems come into being.Based on the business scenario of the M Bus Cloud platform,this paper uses a variety of machine learning recommendation algorithms and combines item attributes and user behavior data to build a recommendation system for the bus cloud.There are two innovative points in this study:1)recommendation systems were integrated into the bus cloud for the first time,and the bus cloud recommendation system was built;2)The time decay function was designed by applying Newton’s law of cooling,and the time transfer function was proposed by integrating the polaris indicators on the basis of the time decay function,which is embedded in the traditional machine learning recommendation algorithms,so that those algorithms can capture the current interest preference of users and the hotness of items in time.The respective average value of click rate and 7-day retention rate are used to compare and analyze the recommendation effect in three stages: before building the bus cloud recommendation system,building the bus cloud recommendation system by traditional machine learning algorithms and building the bus cloud recommendation system by machine learning algorithms with embedded time transfer function.The conclusions are as follows:1)When the bus cloud recommendation system was built with traditional machine learning algorithms,the average value of click rate increased by 8.545% and the average value of 7-day retention rate increased by 5.14% compared with that before it was built,which proves that the construction of bus cloud recommendation system is meaningful;2)The average value of click rate increased by 10.12% and the average value of7-day retention rate increased by 9.4% when the bus cloud recommendation system was built with machine learning algorithms embedded with time transfer function than with traditional machine learning algorithms,which proved that the recommendation effect of machine learning recommendation algorithms with time factor was better than that of traditional machine learning algorithms.