Emerging Trends and Techniques in Machine Learning and Internet of Things-Based Cloud Applications

Abstract

Due to recent development in technology, major changes have been noticed in human being’s life. Today’s lives of human being are becoming more convenient (i.e., in terms of living standard). In current real-world’s applications, we have shifted our attention from wired devices to wireless devices. In result, we moved into the era of smart technology, where many internet devices are interconnected in a distributed and decentralized way. Such internet connected devices (ICDs) or internet of things (IoTs) are generating a lot of data (i.e., via communicating other smart devices). With the tremendous increase in the amount of data, there is a higher requirement to process this huge amount of data (generated through billion of ICDs) using efficient ML algorithms. In the past decade, we refer data mining (DM) algorithms to make some decision from collected data-sets. But, due to increasing data on a large scale, DM fail to handle this data. So, as substitute of DM algorithms and to refine this information in an efficient manner, we require tradition analytics algorithms, i.e., ML or DM algorithms. In current scenario, some of the ML algorithms (available to analysis this data) are supervised (used with labeled data), unsupervised (used with unlabeled data) and semi-supervised (work as reward-based learning). Supervised learning algorithms are like Linear Regression, Classification, and k-nearest neighbor 150(KNN), etc. Whereas, unsupervised learning algorithms are clustering, k-means, etc. In general, ML focuses on building the systems that learn and hence improves with the knowledge and experience. Being the heart of artificial intelligence (AI) and data science, ML is gaining popularity day by day. Notice that a sub-set of AI is ML. Several algorithms have already been developed (in the past decade) for processing of data, although this field focuses on developing new learning algorithm for big data computability with minimum complexity (i.e., in terms of time and space). ML algorithms are not only applicable to computer science field but also extend to medical, psychological, marketing, manufacturing, automobile, etc. On another side, Big Data including Deep learning are the two primary and highly demandable fields of data science. Here, Deep learning is a subset of ML, also a part of computer vision or AI. The large (or massive) amount of data related to a specific domain which forms Big Data (in form of 5 Vs like Velocity, Volume, Value, Variety, and Veracity), contains valuable information related to various fields like marketing, automobile, finance, cyber security, medical, fraud detection, etc. Such real-world’s applications are creating a lot of information every day. The valuable (i.e., needful, or meaningful) information required to be processed (or retrieved) from analysis of this unstructured/large amount of data for further processing of the data for future use (or for prediction). Big organizations have to deal with the large amount of data for prediction, classification, decision making, etc. The use of ML algorithms for big data analytics (DA) includes deep learning, which extracts the high-level semantics from the valuable (meaningful) information form the data. It uses hierarchical process for efficient processing and retrieving the complex abstraction from the data. Hence, this chapter discusses several algorithms of ML, to analysis Big Data. The AI subset, including ML algorithms, is also, Deep learning algorithms is being discussed here (i.e., analyzing this Big Data for accurate prediction). Later, this chapter focuses on the benefits of ML, deep learning algorithms in analyzing the large amount of data (i.e., in unsupervised or unstructured form) for numerous complex problems like information retrieval, medical diagnosis, cognitive science, indexing using semantic analysis, data tagging, speech recognition, natural language processing (NLP), etc. Also, weakness, raised issues, and challenges (during analysis big data) using (in) ML or deep learning have been discussed in detail. In other words, research gaps in using ML, deep learning algorithms for big data will also be discussed (with covering future research directions/trends). In last, the importance of smart era, computational intelligence, AI has been discussed in this chapter in detail.