Multi-Objective Feature Selection for Distributed Systems in Three Stages

This paper focuses on tackling the difficulties of classifying large amounts of data and investigating practical methods to complete the task.The two phases of classification are training and testing (Yousafzai, Hayat, and Afzal, 2020).During the training phase, a classification algorithm uses class labels to identify particular patterns in training data.During the testing phase, a trained classification algorithm predicts a class label for a test dataset.Additionally, big data is a vast collection that continues to grow quickly over time.The magnitude and complexity of this data made it impossible for any of the traditional data processing tools to store or handle.The magnitude of Big Data is a factor in how long it takes the classifier to categorise it and how much memory it uses.Distributed classifier training can potentially process massive data rapidly and effectively.The scalability and efficiency limitations of learning approaches make handling large volumes of data extremely difficult (Thudumu, Branch, Jin, and Singh, 2020).For example, when the computational difficulty exceeds the primary memory, the method cannot scale efficiently because of memory constraints.Distributed learning techniques could be used to scale and handle large data sets.Because distributed learning techniques can divide learning operations across multiple processors, they could make processing large data sets easier.Distributed learning makes use of conventional machine learning techniques.A unique ensemble machine and deep learning approaches are required to improve the precision of distributed learning.Ensemble strategies employ many classification techniques to improve predictive accuracy over a single classification strategy.Therefore, the distributed ensemble machine and deep learning technique were suggested in this paper.According to Kumar, Bashir, Rashid, and Kharel (2021), machine-generated datasets have become larger in recent years.Massive amounts of information that exceed the capabilities of traditional database management systems are comprised of vast quantities of data.Big data is pervasive because modern data-intensive systems can handle a variety of data sources and formats (Erraissi and Belangour, 2018).The necessity for effective partitioning strategies that meet demands for memory use, processing speed, and implementation time has increased as datasets continue to grow.The problem with large data is that