Article Preview
Top1. Introduction
Learning ability is an important part of human intelligence. People can constantly learn new knowledge from past experience or other people's experience, so as to promote the continuous progress and development of human society. Since the advent of electronic computers in the 20th century, people have been pursuing to make machines think and learn like humans. Machine learning has been widely studied. Its purpose is to make computers become human assistants, improve human work rate, and realize the research and exploration of human thinking essence. In the past 20 years, machine learning has been used in network search, spam filtering, recommendation system, advertising, credit evaluation, fraud detection, stock trading and drug design. It has been rapidly popularized in the fields of computer science. Machine learning is an interdisciplinary subject involving probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and so on. It focuses on how computers simulate or realize human learning behavior, so as to obtain new knowledge or skills, reorganize the existing knowledge structure, make it continuously improve its own performance and effectively improve learning efficiency. It is the core of artificial intelligence and the fundamental way to make computers intelligent. The accumulation and rapid growth of all kinds of data bring great challenges to data analysis. As the core technology of data analysis, machine learning extracts rules or knowledge from existing data, so as to provide decision-making basis for human beings in unknown situations. Machine learning model can match the data well, but the adaptability of the model is far less than the complexity of the data. Any model can not ensure the complete matching of various data, and this incomplete matching relationship is often reflected through errors. How to make rational and effective use of errors is a hot issue in the field of machine learning. Statistics is one of the mainstream tools to study data of a certain scale. The difficulties faced by big data analysis highlight the importance of statistics. Therefore, statistical methods are used to study the characteristics of errors, so as to improve the effect of data analysis. The data standardization of matching probability and measurement error is an important part of multi statistical analysis modeling method. Machine learning method highlights the correlation between process variables, eliminates the nonlinear characteristics of the process and the influence of various dimensions, and simplifies the structure of data model. Data standardization based on machine learning usually includes two steps: centralized data processing and non quantitative data processing. On the basis of a large number of sample data, how to scientifically calculate the unknown parameters of events is an important problem in the study of probability and statistics. At present, we need to estimate the matching probability of the matching algorithm, that is, the probability of events(Kim, G.Y., Han,D.S., Lee,Z., et al.(2020)Karim, S., Ye Z., (2017)Laghari, Asif.A., Hui H., (2018)). The statistical method is also known as the statistical method. The error model and compensation method of angle encoder in torsional vibration characteristic measurement system are proposed by traditional reference (Chen, B., Peng,C., Huang,J. et al.(2019)Karim, S., Ye,Z., Shoulin,Y., (2019)Ibrar, M., Jianing,M., Shahid,K., (2018)). According to the principle of spatial error transfer and based on the theory of multi-body system, the geometric error angle measurement error model is constructed. The error matrix of geometric error is constructed by using the benefit Hartenberg method, and the error compensation function is obtained. The influence of geometric error on angle measurement is analyzed. The proposed error compensation method has high operation efficiency, but the conversion process of this method is complex, so it is unable to obtain short-term modeling data, resulting in low operation accuracy of error compensation. For this reason, a modeling method for statistical characteristics of matching probability and measurement error based on machine learning is proposed. The uncertainty of matching probability variables, process, time-varying and transformation time is optimized, and measurement error is counted and corrected. Based on multivariate statistical analysis, combined with the characteristics of process steady-state monitoring and the mutual conversion of various steady-state modes, the statistical characteristic model of measurement error is established. Aiming at the key problems in the multi-mode production process such as data classification, modal identification and feature extraction, the operation accuracy of the statistical model of matching probability and measurement error is improved to ensure the model processing effect.