Research And Application Of Home Decoration Scheme Recommendation Algorithm Based On Collaborative Filtering

Recently, with the advancing development of the Internet technology and the improving level of people's consumption, e-commerce of home decoration in China is being melted into all sides around us. Recommendation engine is similar to the people who recommend products in our real life, which provides great convenience for people who explore the appropriate home decoration scheme during making home renovation and simultaneously brings more value to the website for home decoration.The usage of recommendation engine obviously improve research and application of collaborative filtering algorithm, while most studies about collaborative filtering technology merely value theory of the algorithm, ignoring the finiteness in practical application. Consequently it's necessary to get some insight into users' behavior and the concrete practical situation on website for home decoration to design an appropriate recommendation algorithm on home decoration scheme, which provides decoration design recommendation service for users who desire to make domestic decoration.This paper not only combines the ideas of many collaborative filtering algorithm to design home decoration scheme recommendation algorithm, but also for data collection and processing of home decoration scheme website home's recommendation engine will be described in detail. There're some primary points suspended on the paper as following:(1) After understanding user behavior of the website for home decoration in this paper, the advantage and disadvantage of algorithm on collaborative filtering algorithm and Slope One Algorithm have been studied and analyzed. Besides, the way to improve Slope One Algorithm between the domestic and overseas has been analyzed and compared as well. Gave the following data in a relatively large amount of data and sparse conditions, it will be difficult to obtain a higher quality of the conclusions recommended for the way to improve existing Slope One algorithm. Thus a recommended home decoration scheme algorithm combined user similarity and project similarity to improve Slope One algorithm is proposed. The algorithm mostlyadopt the idea of collaborative filtering, which not only take the project similarity of user-Project Matrix into consideration in project similarity calculation, but also the effect of project itself'ssimilarity.(2)After collection and processingabout a month of the user behavior logs bythe internship company, the data have been divided into training and test of data sets, and compare new algorithmwithslope one algorithm,we verified by MAE value measure proof thatnew algorithm own better recommendation accuracy. Consider the influence of different similarity, we are also test the new algorithm that take different similarity measure to compare verification, it is proved that improvement strategiesof adopts the law of cosines similarity measure better than others.(3)We applied the home decoration scheme recommendation algorithms to icolor home decorationwebsite,andmakerequirementanalysisandarchitecturedesign of recommendation engine, design and implementation the several parts that collection and transportation of user behavior logs, offline data preprocessing and home decoration recommended algorithm, in order to avoid the influence of large amounts of data calculation to website, duringthe procedure of data processing and computing, combine Hadoop, Hive etc.big data related technology and Mahout, taken offline manner to calculate, design to the relevant data in database, Finally, we will make functional testing to recommendation engine of the home decoration website and display the interface, it complete this project originally expected results.