Refining Lazy Learners of Machine Learning Algorithms Facing Concept Drift

In machine learning, concept drift is the gradual change in the relationship between the input and the target. Typically, this could be an unanticipated shift in the way that input and output data relate to one another over time. Usually, a machine learning model is trained and validated its performance across several metrics. The findings are also satisfactory but something unforeseen and unexpected happened with input and the output prediction has gone crazy. It fails and fells victim to a phenomenon called concept drift. It happens because “change is the only constant in life.” Concept drift is now discussed as a strong logic to be considered in the accuracy of machine learning algorithms. The required datasets that are currently available for assessing the working of lazy machine learning algorithms. This paper discusses the detection of “Concept Drift” faced by lazy learners of Machine Learning (ML) algorithms and proposed its solution. Research directions are also examined and discussed. This will immediately benefit researchers in their comprehension of research under the idea of idea of Concept Drift by offering state-of-the-art knowledge. In order to address the shortcomings of lazy ML algorithms such as K-Nearest Neighbor, Logical Weighted Regression, and Radial Base function, a novel heuristic-weight based technique has been employed. XOR (Exclusive OR) is not linearly separable with Radial Base Function which has been implemented to make it linearly separable in Python successfully.