Combination of SOM, SVR, and LMKNN for Stock Price Prediction

Support Vector Machine (SVM) is a strong AI calculation utilized for characterization and relapse errands. One of the shortcomings of SVM is that it doesn't offer direct help for multi-class grouping. The K-Nearest Neighbort (KNN) strategy is known as a basic, quick, and effectively implementable AI technique in different fields, in spite of the fact that its exactness can, in any case, be gotten to the next level. One central disadvantage of KNN is the position of new information into a class in light of most of the neighbors (Vote Larger part Framework), which can be dangerous when the good ways from each closest neighbor vary essentially from the distance of the testing information, frequently prompting misclassifications in KNN. The objective of this exploration is to propose a model that joins Self Organizing Map (SOM), Support Vector Reggression (SVR), and Local Mean-Based K-Nearest Neigbort (LMKNN) to work on the precision of stock cost forecast. From the experimental outcomes, the most elevated exactnesses were acquired for the accompanying stock information: 94.66% for BBCA with the Laplacian piece, 89.29% for ASII with the Gaussian bit, 98.64% for TLKM with the Gaussian part, 98.31% for BBRI with the Gaussian bit, and 96.31% for PGAS with the Gaussian bit.