Big Data Analytics and Its Algorithms

Big data Analytics involves gathering data from several resources, store it in a way that it becomes structured enough to be processed by data analysts and ultimately deliver the data product useful to organizations or businesses. The process of converting large amount of unstructured raw data, gathered from different sources to produce a filtered, clean and useable data bunch forms the purpose of data analytics.

Mining algorithms for certain problems: Along with Big data, big data mining appeared simultaneously concluding that finding anything from big data will be one of the major tasks in research domain. Data mining algorithms play a vital role in big data analysis in terms of computation cost, storage requirement and accuracy results. Some examples of big data mining algorithms are given below

Clustering algorithms: When we talk about big data, traditional clustering algorithms will become absolute as they typically need the data present to be in same format and the data must be loaded on same computer to find information of our need. Many researchers from various fields have been attracted towards the multi-dimensional and large scale dataset and due to that several solutions have been presented in the past few years, the properties of big data still brings up many new challenges for data cluster algorithms. Among them, reducing data complexity is one major issue for big data clustering.

Classification algorithms: Similar to clustering algorithms for big data mining, different researchers also attempted to upgrade the traditional classification algorithms to make them process data mining on a parallel computing environment. In the classification algorithm the input data that are generated by different data sources will be processed by a heterogeneous group of learners. Takin et al. presented a novel classification algorithm called ‘Classify or send for classification (CoS)”. It is assumed that each learner can be used to process the input data in two distinct ways in a distributed data classification system. One to perform a classification function while other forwards the input data to another learner to get them labeled. The information will be continuously exchanged between different learners.

Frequent pattern mining algorithms: Most of the researchers working on frequent pattern mining algorithm (involves association rules and sequential pattern mining) initially focused on analysing large scale data of big shopping mall chains. The number of transactions were more than millions, the issue about how to handle large data was studied for many years. One of the many ways suggested was the FP-Tree, using tree structure to include frequent data patterns to reduce the computation time of mining data. In addition parallel computing and cloud computing technologies attracted researchers.

Map-Reduce solution was used for the studies to enhance the efficiency of Frequent pattern mining algorithm, the study also allowed users to express their specific interest constraints in the algorithm. The performance of data mining capability of these algorithms can be increased by using cloud computing than algorithms running on a single machine.