Concept Drift Detection with Optimal Machine Learning Model for Data Classification
S. Caxton Emerald, T. Vengattaraman · 2022 6th International Conference on Trends in Electronics and Informatics (ICOEI) · 2022
Recently, massive quantity of data streams have been produced rapidly due to the continual technological advancements and it poses concept drifting (CD) as a challenging problem. Since the performance of the classification model gets degraded owing to the presence of concept drifting, it is needed to design effective concept drift detection approaches. Classical classification models are not predicted to discover the pattern in a non-stationary data distribution. In real time scenarios, the classification models are required to properly identify the concept drift and adapt over time. With this motivation, this article develops a new concept of drift detection with chimp optimization algorithm based machine learning (CCD-COAML) model for data classification. The proposed CCD-COAML technique mainly aims to identify the concept drift and then classify the high dimensional data. Initially, the CCD-COAML technique undergoes pre-processing in three distinct stages namely format conversion, data transformation, and chunk generation. Besides, ADaptive WINdowing (ADWIN) approach was implemented to the detection of concept drift and then multilayer perceptron (MLP) model is employed for the data classification process. For improving the data classification efficiency of the MLP model, the weight values can be selected optimally by using COA and thereby improve classification accuracy. A wide-ranging experimental analyses is conducted on benchmark dataset and the outcome stated the remarkable performance of the CCD-COAML methodology on state-of- the-art approaches.