Research And Application On Data Preprocessing System Of Mobile Internet Data

Mobile internet and big data is one of the most interesting issues in recent years.Telecom operators could gain big data when they provide service on mobile internet. How to extract value from this big data efficiently is the key problem to telecom operators and researchers. Data mining is a classic method to solve this problem. There have been a lot of researches on data mining algorithms.Data preprocessing is one of the most important step in data mining. Data preprocessor could provide data with high quality to data mining models, and it world also improve efficiency of data mining models. However, preprocessing of data accounts for a large proportion of the workload in the data analysis work. Especially for such a complex time-series data, preprocessing of data would account for over 80% of the workload.Building automation data preprocessing systems is a necessary way to improve work efficiency of data analysis.There are three problems on building automation data preprocessing system. They are how to automatically select preprocessing algorithms and parameters, how to automatically retrain the preprocessing model and how to evaluate the effect of data preprocessing. This paper provides a heuristic-searching-based portfolio strategy on automatically algorithm combination.This paper includes three research points. First, we research and analysis several of data preprocessing methods and built a data preprocessing systems for mobile internet network data. Then we propose a data preprocessing process for preprocessing of time-series data. Finally we provide a heuristic-searching-based portfolio strategy on automatically algorithm combination.