Welcome to My Machine Learning Journey!

Hey there! I'm Krishna Gupta. I've spent the last six months learning the art of Machine Learning, and my journey is still ongoing.
In this repository, I will share my recipes for different Machine Learning algorithms, starting with the basics and progressing to advanced topics. You’ll also find crisp, short notes to help you better understand the code. ╰(°▽°)╯

Topics Covered:

• Level 1: Data Preprocessing
• Level 2: Regression Models
• Level 3: Classification Models
• Level 4: Clustering Models
• Level 5: Association Rule Learning
• Level 6: Reinforcement Learning

Popular Techniques Explained:

• Fit: This tells the algorithm to learn how different features contribute to an output. It's like teaching a baby how to build a Lego tower using various blocks.

• Transform: Here, the algorithm actually creates new outputs when given a set of features, similar to the final step of building the Lego tower.

• Overfitting: This happens when an algorithm spends too much time learning from the training dataset. As a result, it gets confused when presented with new data.
  Example: It's like drawing a cat too many times from the same reference image. When given a new image, you can't draw it, or your drawing looks nothing like the original.

• Underfitting: The opposite of overfitting, this occurs when the algorithm hasn't learned enough, possibly due to insufficient data or inadequate training.
  Example: It's like not learning enough to draw a cat, because you didn't practice enough or the wrong features were used.

---

Level 1 - Data Preprocessing

X → Independent Variable (Features)

These are the data points used to make predictions.

• Example 1: In predicting house prices, X might include features like the house's dimensions, number of bedrooms, or distance from the city.
• Example 2: In a classification model like animal image recognition, X could represent the weight, height, and color of the animal.

Y → Dependent Variable (Target)

This is the output that the model predicts.

• Example 1: In predicting house prices, Y would be the actual price of the house.
• Example 2: In a classification model, Y could be the category (e.g., DOG, CAT, BIRD, etc.).

The overall goal of an ML algorithm is to establish a relationship between X (independent features) and Y (dependent output).
This relationship can be represented as: [ Y = F(X) + e ] where ( e ) is random noise or error.

Preprocessing Tools:

• Simple Imputer: Fills missing values in your dataset by calculating the average value of the entire column. This helps make missing values more realistic and natural.

• Feature Scaling: Scales data to fit within smaller ranges:

  1. Normalization (range [0, 1]) – Typically used in KNN, neural networks.
  2. Standardization (range [-3, +3]) – Commonly used in linear regression, support vector machines (SVM).

• One-Hot Encoding: Converts categorical variables into numerical form, especially when there’s no inherent relationship between the categories (e.g., CAT, DOG).

• Label Encoding: Converts categorical values into numeric form but is used when there is a relationship between categories, such as customer satisfaction levels (low, medium, high), or binary choices (YES/NO).
  Example: Low < Medium < High, or YES (1) / NO (0).

---

📊 Some Statistical Functions of Pandas Library

1. dataset.head() → Displays the first five rows of the dataset.

   

2. dataset.describe() → Provides statistical information such as count, mean, standard deviation, min, max, and quartiles (1st, 2nd, 3rd, 4th).

   

3. dataset.shape() → Shows the dataset's dimensions (rows × columns).

   

4. dataset['Outcome'].value_counts() → Displays the count of each unique value in the target (dependent) variable.

   

5. dataset.groupby('Outcome').mean() → Groups data based on the Outcome column and calculates the mean for each group.

   

6. dataset.isnull().sum() → Checks for missing values in each column and provides the total count per column.

   

📈 Evaluation Techniques for Regression and Classification

🟢 Classification Metrics

For classification problems where y_test and y_pred are categorical (e.g., classes like 0, 1, 2), common evaluation metrics include:

• Confusion Matrix → Visualizes correct vs. incorrect predictions.
• Accuracy Score → Measures the proportion of correctly classified instances.

🔵 Regression Metrics

For regression problems where y_test and y_pred are continuous values, the following metrics are commonly used:

• Mean Absolute Error (MAE) → Measures the average absolute differences between actual and predicted values.
• Mean Squared Error (MSE) → Computes the average squared differences between actual and predicted values.
• Root Mean Squared Error (RMSE) → Square root of MSE, providing a more interpretable error measure.
• R² Score (Coefficient of Determination) → Indicates how well the model explains the variance in the target variable.