Application Of Machine Learning Algorithms In Data Mining

With the rapid development of the information technology,computer science and the Internet in recent years,various social fields have all ac-cumulated a large scale of data. The data mining technology which studies how to effectively use enormous dataset to find the useful information gets great advances based on the application of information sys-tems,traditional statistical techniques,modern artificial intelligence and corresponding technology of database and statistics.Machine learning is one of the main methods to solve the data min-ing problem.Machine learning is an approach making systems to self-improve and getting the computer program to behave better as the accumulation of experience.Though machine learning can’t make a com-puter reach the learning ability of human beings,it gives a computer the ability to extract features,find latent disciplines from large amount of data and is extensively applied in the data mining field.In this paper,we use machine learning method to solve two concrete data mining problems.The first one is to use the measure report dataset of the mobile terminals’receiving signal to conduct the outdoor localization in the GSM network. For this subject,we put up with a three-phase solu-tion based on support vector machine and k nearest neighbor algorithms which achieve far better precision and less time consuming comparing with traditional methods.The second one is to use the users’ information dataset to solve the link prediction problem in Sina Microblog.For this subject,we propose an effective feature set having several different di-mensions according to the characteristics of Sina Microblog.Then we use an improved support vector machine algorithm to do the model training and link prediction,which reaches good precision and low complexity and compare this method with other classical machine learning algorithms.At last,we do the importance ranking of the proposed feature set and find that the link of users in Sina Microblog is mainly affected by the users’ interest and social association.