Fit and Predict: A Machine Learning Workflow

1. Introduction: What Does “Fit and Predict” Mean?

In machine learning, fitting a model means training it to recognize patterns in data, while predicting involves using those patterns to forecast outcomes. This guide walks through this core workflow using house price prediction as an example.

Key Concepts:

• fit(): Trains the model on data.
• predict(): Generates predictions from new inputs.

---

Step 1: Prepare Your Dataset

Load and Inspect Data

import pandas as pd
data = pd.read_csv('housing_data.csv')

# View structure and missing values
print(data.head())
print(data.isnull().sum())

Key Tasks

• Remove or impute missing values.
• Identify relevant features (e.g., SquareFeet, Bedrooms).

---

Step 2: Fit the Model

Define Features (X) and Target (y)

X = data[['SquareFeet', 'Bedrooms']]
y = data['SalePrice']

Train a Decision Tree

from sklearn.tree import DecisionTreeRegressor
model = DecisionTreeRegressor()
model.fit(X, y)  # Fit the model to the data

What Happens During Fitting:

• The model learns rules like:

if SquareFeet > 2000 and Bedrooms == 3 → predict $420,000

Step 3: Make Predictions

Generate Price Estimates

new_data = [[1800, 2], [2400, 4]]  # New houses to predict
predictions = model.predict(new_data)
print(predictions)  # e.g., [320000, 475000]

Evaluate Accuracy with MAE

from sklearn.metrics import mean_absolute_error
train_predictions = model.predict(X)
mae = mean_absolute_error(y, train_predictions)
print(f"Training Error: ${mae:.2f}")

Step 4: Validate and Improve

Avoid Overfitting with a Validation Set

from sklearn.model_selection import train_test_split
X_train, X_val, y_train, y_val = train_test_split(X, y, test_size=0.2)

model.fit(X_train, y_train)
val_mae = mean_absolute_error(y_val, model.predict(X_val))
print(f"Validation Error: ${val_mae:.2f}")

Upgrade to a Random Forest

from sklearn.ensemble import RandomForestRegressor
improved_model = RandomForestRegressor(n_estimators=100)
improved_model.fit(X_train, y_train)

FAQ: Common Fitting Challenges

Q: Why does my model perform poorly on new data?

Overfitting: The model memorized noise in the training data.

Fix: Simplify the model or use cross-validation.

Q: How much data do I need to fit a model?

• Start with at least 100-1000 samples per feature. Smaller datasets benefit from algorithms like decision trees.

Q: What if predictions are nonsensical?

• Check for: -- Data leaks (e.g., target variable accidentally included in features). -- Incorrect feature scaling.

Fitting and predicting is the core of machine learning—train models, validate, and refine for accurate results.