This paper adopts three data methods: multiple choice research, cluster analysis and weight research.

1. Multiple choice research: Multiple choice analysis can be divided into four types, including multiple choice, single choice-multiple choice, multiple choice-single choice and multiple choice-multiple choice.

2. Cluster analysis: Cluster analysis classifies sample objects based on multiple research titles. If clustering is based on samples, the system will automatically identify whether to use K-means clustering algorithm or K-prototype clustering algorithm by using the "clustering" function in the advanced method module of SPSSAU.

3. Weight research: Weight research is used to analyze the importance of various factors or indicators in the comprehensive system and finally build a weight system. There are many methods to study the weight, including factor analysis, entropy method, AHP analytic hierarchy process, TOPSIS method, fuzzy comprehensive evaluation method, grey correlation method and so on.

First, regression analysis.

In practical problems, it is often encountered that several variables need to be considered at the same time, such as the relationship between height and weight, blood pressure, age and so on. The relationship between them is very complicated and cannot be studied accurately, so the relationship between them cannot be expressed in functional form. In order to study the relationship between these variables, it is necessary to obtain data through a large number of experimental observations and use statistical methods to find out their relationships, which reflect the statistical laws between variables. One of the statistical methods is regression analysis.

The simplest is linear regression, which only considers the relationship between a dependent variable Y and an independent variable X. For example, if we want to study the relationship between people's height and weight, we need to collect a large number of different people's height and weight data and then establish a linear model. Next, the unknown parameters need to be estimated, and the least square method can be used here. Finally, the regression equation should be tested for significance to verify whether y changes linearly with X. Here we usually use t test.

Second, analysis of variance.

In practical work, there are many factors that affect one thing, and people hope to observe the influence of various factors on the experimental results through experiments. Variance analysis is a mathematical statistical method to study whether the change of one or more factors has a significant impact on the observed values of experimental results, so as to find out better experimental conditions or production conditions.

The quantitative indicators observed by people in experiments are called observed values, the conditions that affect the observed values are called factors, the different states of factors are called levels, and a factor may have multiple levels.

In an experiment, a series of different observation results can be obtained, some of which are caused by different treatment methods or conditions. This phenomenon is called factor effect. Some are caused by errors, which are called experimental errors. The main work of variance analysis is to decompose the total variation of measured data into factor effect and experimental error according to the different reasons that cause the variation, and make a quantitative analysis of it, and compare the importance of various reasons in the total variation as the basis for statistical inference.

For example, we have four components produced by different formulations, and we want to judge whether there are significant differences in their service life. Here, the formula is the factor that affects the service life of parts, and four different formulas become four grades. It can be judged by analysis of variance.