A fast extraction method of online catering master label data based on Gaussian estimation

Abstract

The invention discloses a Gaussian estimation-based online catering main tag data quick extraction method. The method comprises the following steps of 1) obtaining store tag data and user score and review data, and preprocessing the store tag data; 2) for each user, calculating a mean value and a variance of Gaussian distribution of a score data set of each tag, namely, score Gaussian distribution of the user under the tag; 3) for each user, performing standardization processing on each tag score of a store each time and the score Gaussian distribution of the tag, and calculating maximum likelihood estimation as a target tag of the user going to the store at that time; and 4) testing an estimation tag, an actual complete tag and the review data, and taking a relative deviation of a matching rate of the tag and the review data as a final evaluation score of a model. According to the method, a maximum likelihood taste tag under the Gaussian distribution is extracted as a mainly selected taste tag of a dining behavior of the user; and the method is relatively high in extraction precision and relatively low in algorithm complexity, and is suitable for actual application scenes.

Description

technical field [0001] The invention relates to the field of data mining and recommendation systems, in particular to a method for quickly extracting main label data of online catering based on Gaussian estimation. Background technique [0002] The data collected in data mining often have various noises, such as missing data or abnormal data. Obviously, noisy data can affect the performance of subsequent modeling. Data denoising is a very important preprocessing step in order to extract data that retains the maximum amount of information. In the process of user data analysis, sometimes using a good data denoising method to improve accuracy is much better than complex algorithm optimization. [0003] The main purpose of designing a recommendation system is to predict user behavior preferences, and analysis materials often come from user historical behavior data. To discover the content of a user's purchase behavior, a common method is to analyze user comments through natura...

AI Technical Summary

Problems solved by technology

[0004] In order to overcome the inability of the existing catering data extraction methods to take into account both content integrity and utility, the present invention provides a Gaussian-based estimation method that has a balanced performance in content integrity and utility after denoising data. Fast extraction method of online catering master label data

Images