A direction is a trend that the market follows throughout a specific time period. Trends can be both upward and downward, corresponding to bullish and bearish markets. Direction detection is an excellent way to predict how the market will move in the future, and it can assist investors in shifting the odds in their favor while trading. Understanding the direction of the stock market is of the utmost importance for any approach that investors take in the market. An investor could suffer substantial losses if invested in a stock without knowing its direction and it happens to be in a prolonged slump. Investor may buy high and sell low if they do not grasp the stock's history and current direction. There are many algorithms available to investors that can pinpoint the exact closing price of any stock without revealing the direction of the closing price. This study uses a variety of machine learning classification methods to try and forecast the direction of the given stock. This study's goal is to create the best models possible by combining several modelling techniques and investigating innovative ways to reduce direction prediction mistakes. The dataset includes daily closing prices of a stock for the previous 22 years, and different models are employed to predict direction changes based on 2% and 4% differences in percentage change in close price. The rule classifies these changes as either positive or no change. To enhance prediction accuracy, technical analysis indicators like Average True Range (ATR) and Volume Weighted Average Price (VWAP) are also integrated as feature variables. The class Imbalance problem is solved using the SMOTE over-sampling technique. Multiple classification models are developed to assess their predictive accuracy, with the Random Forest (RF) model showing the highest accuracy of 91% for 2% variation and 96% for 4% variation in closing price. Also, Neural Networks provided the next best results in predicting stock price direction with 85% accuracy for 2% variation and 95% accuracy for 4% variation. The key takeaway from this is the significant utility of diverse classification modeling techniques in effectively forecasting the direction of closing prices for the stock in question.