The growing complexity and speed of financial markets indicate the need for research in high-frequency trading, with decision-making processes being timely and informed so as to maximize profitability. This paper explores the use of machine learning models to enhance market efficiency by High-Frequency Trading (HFT) through price prediction in the short run, trade signals generation, and market trend detection. The strengths of three models, namely Long Short-Term Memory (LSTM), Gradient Boosting Machines (GBM), and Artificial Neural Network (ANN) are gauged against market movements for predictive decisions on trades, and research is done in machine learning. The results of these models are measured on all grounds, including better performance over others at shorter price predictions and signal generations about the trading actions along with their respective trends of market changes. Specifically, LSTM had a prediction accuracy of 97.5%, trade signal accuracy of 96.2%, and market trend detection accuracy of 98.0%. GBM came next with good performance, scoring 93.2% in price prediction, 91.8% in trade signal generation, and 94.5% in market trend detection. The ANN model, though less effective, still contributed valuable insights with accuracy rates of 85.7%, 84.4%, and 86.3%, respectively. Further results show the strong impact that can be produced on decisions and profit in high-frequency trading using the machine learning models, especially the LSTM model. In addition, this study draws attention to real-time data requirements, including both sentiment analysis and macroeconomic variables, for enhanced model performance that can be tuned in real-time to reflect variations in market dynamics.