PHASE B: Machine Learning Basics (3-4 Weeks)

This phase is short but critical.
These concepts power almost every applied AI pipeline.

What to Learn

• supervised learning
• classification
• regression
• train/test split
• overfitting and underfitting
• evaluation metrics

Core tool:

• scikit-learn

Week-by-Week Plan

Week 1

• ML workflow basics
• train/validation/test split
• linear regression and logistic regression

Week 2

• tree-based models
• feature preprocessing
• confusion matrix, precision, recall, F1

Week 3

• model comparison
• cross-validation
• overfitting controls (regularization, depth limits)

Week 4 (Optional but Recommended)

• hyperparameter tuning
• error analysis
• model report writing

Build These 3 Projects (Must Do)

Project 1: Spam Detection

Goal:

• classify text messages as spam or not spam

Skills:

• text vectorization (CountVectorizer, TfidfVectorizer)
• classification metrics

Minimum deliverables:

• notebook with baseline and improved model
• confusion matrix + precision/recall analysis

Project 2: Movie Sentiment Analysis

Goal:

• classify reviews as positive/negative

Skills:

• preprocessing pipeline
• logistic regression / naive bayes baseline
• model error analysis

Minimum deliverables:

• model comparison table
• examples of wrong predictions and why

Project 3: Tabular Prediction Model

Goal:

• predict a numeric or categorical target from tabular data

Skills:

• handling missing values
• feature engineering basics
• regression/classification baseline

Minimum deliverables:

• clean feature pipeline
• validation score and model interpretation notes

Scikit-Learn Starter Pattern

from sklearn.model_selection import train_test_split
from sklearn.pipeline import Pipeline
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.linear_model import LogisticRegression

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

clf = Pipeline([
    ("tfidf", TfidfVectorizer()),
    ("model", LogisticRegression(max_iter=1000))
])

clf.fit(X_train, y_train)

Why These 3 Projects Matter

These are simple, but they train the most important habits:

• turning messy data into model-ready inputs
• evaluating beyond accuracy
• reporting results clearly
• improving models systematically

Next Step

After this phase, move to:

• NLP Foundation Roadmap: Transformers, Hugging Face, and Research Portfolio