Machine Learning Projects

This workspace contains two small ML projects:

1. Iris classification model (Random Forest) in irisModel/
2. Smartphone price regression model (Random Forest) in smartphoneModel/

It also includes a synthetic data generator for smartphone pricing.

Requirements

• Python 3.8+
• pandas
• numpy
• scikit-learn

Install dependencies:

pip install pandas numpy scikit-learn

Project: Iris model

Location: irisModel/

Dataset: irisModel/Dataset/

Notebook: irisModel/model.ipynb

What it does:

• Trains a Random Forest classifier on train.xlsx.
• Predicts on test.xlsx.
• Writes results to result.xlsx and prints accuracy stats.

How to run:

Open irisModel/model.ipynb and run the cells top-to-bottom.

Accuracy reporting (Iris):

After generating result.xlsx, use this snippet to report accuracy:

from sklearn.metrics import accuracy_score

accuracy = accuracy_score(result['actual'], result['predictions'])
error_percentage = (1 - accuracy) * 100

print(f"Accuracy: {accuracy * 100:.2f}%")
print(f"Error Percentage: {error_percentage:.2f}%")
print(f"\nNumber of correct predictions: {(result['predictions'] == result['actual']).sum()}")
print(f"Number of incorrect predictions: {(result['predictions'] != result['actual']).sum()}")

Latest results (Iris):

• Accuracy: 97.06%
• Error Percentage: 2.94%
• Correct predictions: 33
• Incorrect predictions: 1

Project: Smartphone price model

Location: smartphoneModel/

Dataset: smartphoneModel/Dataset/

Notebook: smartphoneModel/model.ipynb

What it does:

• Trains a Random Forest regressor on Dataset/train.csv.
• Tests on Dataset/test.csv.
• Saves predictions to Dataset/result.csv and prints MAE, RMSE, and R2.

How to run:

Open smartphoneModel/model.ipynb and run the cells top-to-bottom.

Accuracy reporting (Smartphone regression):

After generating Dataset/result.csv, you can compute accuracy as the fraction of predictions within a tolerance of the true price:

tolerance = 0.10  # 10%
within_tol = (result['prediction'].sub(result['actual']).abs() <= tolerance * result['actual'])
accuracy = within_tol.mean()
error_percentage = (1 - accuracy) * 100

print(f"Accuracy (within {tolerance*100:.0f}%): {accuracy * 100:.2f}%")
print(f"Error Percentage: {error_percentage:.2f}%")
print(f"\nNumber of correct predictions: {within_tol.sum()}")
print(f"Number of incorrect predictions: {(~within_tol).sum()}")

# Exact-match accuracy (usually near 0 for continuous targets)
exact_match = (result['prediction'].round().astype(int) == result['actual'])
print(f"\nExact-match Accuracy: {exact_match.mean() * 100:.2f}%")
print(f"Exact correct: {exact_match.sum()}, exact incorrect: {(~exact_match).sum()}")

Latest results (Smartphone regression):

• Accuracy (within 10%): 40.00%
• Error Percentage: 60.00%
• Correct predictions: 80
• Incorrect predictions: 120
• Exact-match Accuracy: 0.00%
• Exact correct: 0
• Exact incorrect: 200

Smartphone dataset generator

Script: smartphones-Data.py

What it does:

• Generates a synthetic smartphone dataset with 1,000 rows.
• Writes the dataset to smartphone_data.csv.
• Prints the first few rows to the console.

How to run:

python smartphones-Data.py