A Survey on Data Collection for Machine Learning: A Big Data - AI Integration Perspective

Abstract

Data collection is a major bottleneck in machine learning and an active research topic in multiple communities. There are largely two reasons data collection has recently become a critical issue. First, as machine learning is becoming more widely-used, we are seeing new applications that do not necessarily have enough labeled data. Second, unlike traditional machine learning, deep learning techniques automatically generate features, which saves feature engineering costs, but in return may require larger amounts of labeled data. Interestingly, recent research in data collection comes not only from the machine learning, natural language, and computer vision communities, but also from the data management community due to the importance of handling large amounts of data. In this survey, we perform a comprehensive study of data collection from a data management point of view. Data collection largely consists of data acquisition, data labeling, and improvement of existing data or models. We provide a research landscape of these operations, provide guidelines on which technique to use when, and identify interesting research challenges. The integration of machine learning and data management for data collection is part of a larger trend of Big data and Artificial Intelligence (AI) integration and opens many opportunities for new research. TaskApproach Techniques Data discovery Sharing Collaborative Analysis [9]-[11] Web [12]-[17]

A Complete Research on Techniques & Technologies of Big Web Data Preparation to Web User Usage Behaviour

“…As a future work, Hadoop approach able to concentrates comprehensively on both the stages: distributed data storage & parallel processing of weblog data and to leverage the strengths of techniques and technologies of individual stages [11,16,21]. In addition, the comprehensive approach is planned to test with different weblogs that cover a large spectrum of various applications, such as, web usage analysis for improvements in fraud detection, product analysis and customer segmentation.…”

“…However, in the present paper, the authors pay an attention only on remarkable findings of techniques and technologies involved in preparing the big web data suitable to analytics. In this direction, some of the authors [10,11,12,14,27] described various methodologies discovered by both research and industry community to pre-process the weblog data efficiently in Big Data environment. According to the research works made by the authors [10,20,27,31] are endorsed that big data storage, big data cleansing, unique user identification, session identification and so on are important and crucial tasks in the big data preprocessing model.…”

“…The research papers [4,11,16,17,18,28] analyze the challenging issues of Hadoop echo-system in dealing with different kinds of data. The authors find-out the necessity of paying an attention by the present researchers for the development of more innovative techniques by makes use of machine learning paradigm.…”

“…Only the records having 200 as the status code is treated as relevant data. Many authors in the literature [10,11,12,14,20,27] are also prominently expressed that removal of failure requests, irrelevant file requests, inappropriate access method requests, web robots requests, internal dummy connection requests and irrelevant log fields to extract more relevant data for the analysis of weblog.…”

Reliability of Training Data Sets for ML Classifiers: A Lesson Learned from Mechanical Engineering

“…Option 6 -Various models of data collections for the purpose of running ML algorithms [33] could also be investigated. However, the sensitivity of this problem domain and the fact that we need the highest level of classifier's accuracy upon the training data set might not open a door to data from shared sources.…”