The Design And Implementation Of Real Time Marketing System Based On Scenario Prediction

With the rapid development of the Internet economy,the physical enterprises have been dying out in recent years.At present,the way of real economy propaganda is still mainly to issue leaflets and advertisements.These propaganda methods can not be more accurately pushed to the audience,and the ultimate effect of propaganda is often unsatisfactory.In order to stimulate the store economy better,it is a good way to use recommendation system to propagate,but most of the current recommendation systems are used for the Internet economic scenarios,not very suitable for the entity consumption scenarios,because the real economy is far more sensitive to geographical location and other situational factors than Internet scenario.In addition,the recommended scheme of the current recommendation system is basically configured by the platform,merchants can not operate on it.To sum up,this thesis designs a recommendation system which pays more attention to the geographical location and other situational factors,and opens the operation authority of the recommendation scheme to merchants.In this thesis,a real-time marketing system based on scenario prediction is designed for the scenario of consuming in the shop.The main functions of this system include: merchants use this system as a marketing tool to launch marketing recommendation scheme,the cosumer's browsing event triggers the system to process in real time and make real-time response.The main contents and achievements of this paper are as follows: 1)This thesis analyzes the requirements of the real-time marketing system based on scenario prediction and gives the design and implementation scheme of the system,the system uses Kafka as message queue to receive user behavior events,Flink and Spark as system computing framework to process data,and Spark is used for data pre-processing,Flink is used for real-time processing,system uses TiDB as data warehouse and Redis as cache;2)Scenario prediction is introduced into the process of traditional collaborative filtering recommendation algorithm to improve the accuracy of recommendation algorithm;3)consumption cycle is introduced to modify user profile preference score,so that user preference score changes with time;4)Finally,the function test and performance test of the system are carried out.Function testing is the completion of functional points in the demanding analysis.The performance testing is to test the real-time and recommendation efficiency of the system.The real-time performance of the system is guaranteed by real-time processing and response to user behavior.The recommendation algorithm efficiency is more accurate than the traditional collaborative filtering recommendation algorithm.The test results show that the system can complete the response within 100 ms,and the accuracy of the recommendation algorithm is improved about 10% compared to the traditional collaborative filtering recommendation algorithms,which verifies the real-time performance and the high efficiency of the system.