STOCK PRICE PREDICTION using Machine Learning📈 | Machine Learning Projects | GeeksforGeeks

---

---

AI Summary

Session Summary:

• Topic: Implementation of KNN algorithm
• Previous Session Recap:
  • Discussed mathematical intuition behind KNN for regression and classification.
  • Advised to watch the previous session for better understanding.
• Agenda:
  • Define the problem.
  • Step-by-step coding demonstration using a stock prediction dataset.
• Dataset:
  • Source: Quandl (financial data extraction).
  • Features: Date, Open, High, Low, Last, Close, Total Trade Quantity, Turnover.
  • Objective: Predict the ‘Close’ value and decide whether to buy or sell stock.
• Libraries Used:
  • Pandas, NumPy, Matplotlib, Quandl.
• Tasks:
  1. Predict the ‘Close’ value (Regression Problem).
  2. Decide whether to buy (+1) or sell (-1) stock (Classification Problem).
• KNN Implementation:
  • Demonstrated KNN for both regression and classification.
  • Used features like ‘Open - Close’ and ‘High - Low’ for predictions.
  • Explained the process of converting stock data into a classification problem.
  • Discussed the importance of choosing the right ‘K’ value using GridSearchCV.
• Model Training and Testing:
  • Split data into training and testing sets.
  • Used KNeighborsClassifier and KNeighborsRegressor from sklearn.
  • Evaluated model accuracy and error metrics (RMS error).
• Results:
  • Classification: Training accuracy (68%), Test accuracy (51%).
  • Regression: Showed actual vs predicted ‘Close’ values, noted high RMS error due to feature selection.
• Conclusion:
  • KNN can be used for both regression and classification tasks.
  • The choice of features and the value of ‘K’ are crucial for model performance.
  • Future sessions will explore other models like Random Forest and LSTM.
• Next Steps:
  • Discuss other topics and models in future sessions.
  • Address any doubts in the comment section.
• End of Session.