An Efficient Optimal Feature Selection-based Multi-Task Learning Model using Dilated Convolutional Neural Network with Improved Doctor and Patient Optimization

Multitask Learning (MTL) is attracting a lot of interest because it may transfer information across projects and increase teamwork by leveraging common frameworks throughout tasks to collaboratively construct an improved model for every assignment. Most conventional MTL approaches only concentrates on choosing features across various tasks, concern on enhancing the limited number of learned parameters. MTL is rigorously demonstrated to increase extrapolation ability given a series of associated duties. Most MTL algorithms presume that every assignment is connected. Nevertheless, if each of the tasks is unrelated, there is a detrimental flow of data between them, and the effectiveness of the classic MTL process suffers. In this paper, the optimal feature selection approach is offered to perform the MTL. Research work is initiated by collecting the data from the large dataset. From this collected data, optimal feature selection is performed using Improved Doctor and Patient Optimization (IDPO). The multi-objective functions like correlation coefficient, relief score, and chi-squared statistic are considered by the IDPO for enhancing the performance of the optimal feature selection process. Lastly, the optimally selected features are transferred to the Dilated Convolutional Neural Network (DCNN) for performing the MTL. The MTL accompanied by the large dataset and solves the multiple struggles. The existing approaches are compared with the results of the developed feature selection methodology to confirm its effectiveness.