Introduction

Machine Learning is learning from data and make decisiomn without being explicitly programmed. It is about creating predictive models from data by finding patterns.

Main phases of Machine Learning

• Data Collection: Gathering data from various sources
• Data Preprocessing:
• Model Training:
• Model Evaluation:
• Model Deployment:
• Model Monitoring:

How Machine Learning Works

ML Models extract pattern from historical data to make predictions or decisions about new/unseen data.

Machine Learning Models

There are 2 main phases

1. Training
2. Inferencing

Machine Learning Types

Multiple types of machine learning, and use the model depending on what you’re trying to predict.

• Supervised Learning
  • Regression
  • Classification
    • Binary Classification
    • Multiclass Classification
    • Multi-label Classification
• Unsupervised Learning
  • Clustering
  • Dimensionality Reduction
  • Association Rules
• Reinforcement Learning
• Semi-Supervised Learning
• Self-Supervised Learning

Supervised Learning

If a model learn from labeled data (Correct answers already know during training).

Ex:

• Speech recognisation: from audio to text transcript
• Tranlation: english to dutch
• Spam filtering: emaill spam or not

Algrithem Types

• Classification: Predicts which category something belongs to (small number of possible outputs).
  • Binary classification: Predicts one or two possible outcomes. Ex: patient is diabetic or not. whether a mail is spam or not.
  • Multiclass classification: Predicts one or multiple possible outcomes. Ex: Movie is horror, comedy or fiction etc.
• Regression: Predicts a continuse numaric value (Many possible outputs and also often many inputs required). House pricing, stock price, sales predictioon etc.

Unsupervised Learning

If a model learns from the data without labeled and try to automatically group them into clusters.

Algrithem Types

• Clustering/Grouping - Group similar datapoints to gether. Ex: Google news - looks for thosends of news articles and group related stories together, based on customer age, salary, location predicting a premium, budget/medium customer.
• Dimensionality Reduction - Take big data set and compress it to much smaller data set using few numbers (While loosing as less data as possible). Ex: Customer data contains lot of information, reduce to key 10 features related to financial.
• Anomaly Detection - Find unusual data points. Ex: Financial fraud systems, AIops etc

Supervised Learning - Training the model

Regression

Teaching a machine learning model to predict a continuous numeric value from input data.

Typical workflow:

1. Collect data
2. Split data
   • Training data
   • Testing data
3. Train the regression model
4. Evaluate using metrics: To check if the model is good, data scientists use evaluation metrics, for example:
   • MAE (Mean Absolute Error) – average prediction error
   • MSE (Mean Squared Error) – squared prediction error
   • RMSE (Root Mean Squared Error) – square root of MSE
   • R² (R-squared) – how well the model explains the data
5. Improve the model

Real world scenario - Iterative training of training, validating & evaluating model.

Classification

• Binary Classification
• Multiclass Classification

1. Binary Classification

Binary classification is a type of supervised machine learning task where a model learns from labeled data to classify inputs into one of two possible categories.

These categories are usually represented as:

• 0 or 1
• True or False
• Yes or No
• Positive or Negative

Popular algorithms for binary classification:

• Logistic Regression
• Decision Trees
• Random Forest
• Support Vector Machines (SVM)
• Neural Networks
• Gradient Boosting (XGBoost, LightGBM)

2. Multiclass Classification

Unsupervised Learning - Training the model

Clustering