Research on Agricultural Products Recommendation System Using Spark Technology

With the rapid development of Internet, the marketing model of agricultural products is constantly innovating. The invention of recommendation system algorithm has brought great changes to Internet marketing promotion, but there are some shortcomings, such as low recommendation speed and the recommendation system does not consider the time factor, which need to be targeted at different scenarios. Internet data grows at EB level every day, and the application of general recommendation system can't meet the needs of storage and calculation at all. The system adopts collaborative filtering recommendation algorithm for agricultural products, runs in Ububtu operating system and is realized by big data Hadoop+Spark programming. The collaborative filtering recommendation algorithm module in SPARK's MLib data mining module is used, and the user-based collaborative filtering algorithm in three types of collaborative filtering is adopted. In view of the certain shelf life of agricultural products, the time decay function should be added to improve the recommendation of agricultural products. The system operation shows that the system conforms to the ISO/lEC 25010 software quality standard, and has good functional applicability, performance efficiency, compatibility, availability, reliability, safety, maintainability and portability. The experience proves that the speed of the improved recommendation system running under the Spark framework has increased by 2.30 times, and the recommendation speed has been obviously improved.