Машинное обучение: регрессия и предсказание данных на Python

Dive deep into linear regression for energy consumption forecasting! This course takes you through the complete process, mirroring a real Kaggle competition.

The course is structured into two parts, covering both theory and hands-on application.

Part 1: Data Preparation & Foundational Models

• Understanding Energy Forecasting Tasks: Learn about different types of energy prediction problems and how to define them effectively.
• Machine Learning Models for Prediction: Explore various machine learning models to minimize prediction error.
• Fundamentals of Machine Learning: Build a solid understanding of the core concepts behind machine learning model construction.
• Key Metrics & Simple Models: Master essential metrics and learn to implement fundamental models like:
  • Linear Regression: The bedrock of regression analysis.
  • Polynomial Regression: Modeling non-linear relationships.
  • Regularized Linear Regression: Techniques like Ridge and Lasso for improved model performance.

Part 2: Practical Application & Optimization

• Data Analysis (ETL) with Pandas: Learn to efficiently load, clean, and combine datasets using the powerful Pandas library.
• Exploratory Data Analysis (EDA): Discover hidden patterns and dependencies within your energy data.
• Linear Regression with Scikit-learn (sklearn): Implement and utilize linear regression models using the widely-used sklearn library.
• Data Interpolation & Extrapolation: Predict values outside the range of your observed data.
• RMSLE Metric Calculation: Understand and calculate the Root Mean Squared Logarithmic Error, a key metric for evaluating linear regression models.
• Linear Regression Optimization: Fine-tune your models by:
  • Selecting optimal parameters.
  • Tuning hyperparameters for best performance.
  • Optimizing memory usage when handling large datasets.
• Ensemble Methods for Enhanced Prediction: Combine multiple linear regression models to improve forecast accuracy.
• Data Export & Import: Seamlessly manage data flow with export and import functionalities, including intermediate results.
• Kaggle Competition Ready: Prepare and export your results in the format required for submitting to a Kaggle competition.

This course provides a comprehensive understanding of using linear regression to solve real-world energy forecasting problems, equipping you with the skills to excel in data science competitions and practical applications.