Research And Application Of Recommendation Technology Based On Logistic Regression

As an important part of modern e-commerce platform, the performance of therecommended technology determines the performance of electronic business platform.Advances in technology brought about the rapid development of the Internet; it will alsobe numerous and various information to show to us. Faced with so much information,we will think about: how to find their interest or useful information. The entire actualsituation: want to discover for their valuable fragments undoubtedly become verydifficult, even full browser again is unrealistic. Personalized recommendation systemdevelopment began in the1990s, the U.S. Conference on Artificial Intelligence, wasattended by Carnegie Mellon RobertArmstrong et al proposed a personalized navigationsystem: WebWatcher. From Stanford University, MarkoBalabanovic et al then also atthe meeting put forward the idea of personalized recommendation system-LIRA. Pulledfrom the development of personalized recommendation system prelude. Content-basedrecommendation system from the very beginning to the later development ofcollaborative filtering recommendation system experienced a short period of20years,and now with the improvement of our mutual network infrastructure, the expansion ofthe network bandwidth, the recommended system the development of more rapid. Therecommendation system development experience is not a very long time, but its impactis huge. In this process has appeared in a variety of recommendation algorithms, themost important representatives: collaborative filtering algorithms, content-basedrecommendation algorithm and recommendation algorithm based on network structure.They appear, as well as the development of an extension, pave the way for the futuredevelopment of the recommended system played a role.In order to study the effect of characteristic attributes played in therecommendation system, we think of a logistic regression model. Logistic regressionlogistic regression analysis, the general application pandemics learn more. The contentof this study are:First, the focus of our research is the recommended algorithm module thebackground recommendation algorithm, real-time product collection filter out the recommended products of interest to the user. Model the scene by a logistic regressionmodel to analyze data sets, and select the best features from the analysis results.Logistic regression function parameters can be self preferably, i.e. to enter a series ofparameters, greater impact on the characteristic parameter for the data sets obtained bythe last training.Secondly, logistic regression models predict the effect of the scene. Sceneprediction analysis in two steps, the first step is to study the parameter selection effect inthe offline case, that is, the establishment of the training parameters of the regressionmodel engine; second step, the application of optimized parameters to test the effect ofonline testing station, i.e. the application to the recommendation system, and analyzethe results.Finally, the study uses logistic regression scene engine technology in the effect ofthe recommendation system, the different data sets. The behavior of the users to buyinto the behavior of the self-behavior and commodities. Our study between therespective characteristics embodied in user interest, based on the recommendation of theprevious methods and cannot be pointed out that the impact of the differentcharacteristics of different e-commerce environment. Mining and commoditycharacteristics, at the same time to study the impact of the different characteristics ofdifferent data sets can be explained by reasons of recommended items to better reflectthe user’s personal preferences.By this method and the standard collaborative filtering recommendationtechnologies, we find that: the former is a certain improvement in the precision andrecall rate recommended, and F1indicators also increased, in time better than thestandard collaborative filtering recommended system. Finally, as the main support tothis theory the practical application of the design and implementation, and operatingresults show, movie recommendation to observe the actual effect of repeated use, thebetter the performance of the application, to achieve the expected results of this paper.Function slightly single, need to be further enriched.