Improving the Speed of Machine Learning Algorithms using Bio-Inspired Techniques

Today, digital data is exploding at a breakneck speed. Traditional data analytics techniques, unfortunately, are rapidly losing their capabilities and efficiencies when dealing with large datasets. This problem has prompted several researchers to develop more effective, efficient, and fast big data analytics tools. Machine Learning (ML)-based approaches are among the most dependable techniques used to extract usable insights from large datasets. However, some of them cannot efficiently handle large datasets, and their training time grows with dataset size. This paper presents two Nature-Inspired techniques for improving the training time of ML algorithms and the processing time of big datasets. The techniques are evaluated on four ML algorithms and large or medium-scale datasets. Results show that the training time of the four ML algorithms was reduced without a significant drop in classification accuracy. Moreover, the proposed methods are significantly faster than two well-known instance selection methods. Furthermore, statistical analysis reveals that the techniques reduced data size significantly, making them suitable for processing large datasets.