Enhancing Stock Price Prediction: Improvising in KNN

Abstract

Stock price prediction is very crucial for informed investment decisions, involving forecasting futurestock values which are based on various factors. K-nearest neighbors (KNN) is a machine learningalgorithm that can assist in predicting stock prices by identifying patterns and similarities betweenthe target stock and its neighboring data points in a multidimensional feature space. However,traditional KNN algorithms encounter challenges like sensitivity to irrelevant features and outliers,potentially compromising predictive accuracy. To address this, integrating Density-Based SpatialClustering of Applications with Noise (DBSCAN) before KNN proves effective. DBSCAN identifiesand filters out noisy data points and outliers, refining the dataset for subsequent KNN analysis. Thisintegration not only mitigates traditional KNN issues but also uncovers underlying data structures,improving overall predictive power in stock market analysis.