/ML-projects

Primary LanguageJupyter Notebook

This repository contains the following ML projects:

1. Water Potability Prediction

The water_potability.py script. This project involves predicting the potability of water based on various features.

  • Purpose: Predict if water is safe for consumption.
  • Usage: Instructions on how to run the script.
  • Dependencies: List of required libraries.

2. Employee Turnover Prediction

The hr.py script. This project involves predicting employee turnover rates.

  • Purpose: Predict the likelihood of employees leaving the company.
  • Usage: Instructions on how to run the script.
  • Dependencies: List of required libraries.

3. Sales Prediction Based on Ad Spend

The ads.py script. This project involves predicting sales based on advertising spend.

  • Purpose: Forecast sales figures based on ad expenditures.
  • Usage: Instructions on how to run the script.
  • Dependencies: List of required libraries.

4. Diabetes Prediction Model

The diabetes.py script. This project involves predicting diabetes based on health metrics.

  • Purpose: Predict the likelihood of diabetes.
  • Usage: Instructions on how to run the script.
  • Dependencies: List of required libraries.

Water Potability Prediction

This project aims to classify water samples as potable (drinkable) or non-potable using machine learning models. It compares the performance of different models using the Receiver Operating Characteristic (ROC) curve.

Overview

The project builds two main models:

  • Logistic Regression
  • Support Vector Machine (SVM)

The performance of these models is evaluated using ROC curves and their respective Area Under the Curve (AUC) scores.

ROC Curve Analysis

The ROC curve is used to evaluate the performance of the classification models by plotting the True Positive Rate (TPR) against the False Positive Rate (FPR). A better performing model will have a curve that hugs the top-left corner of the graph, indicating high sensitivity with a low false positive rate.

In this project, we compared the performance of Logistic Regression and SVM models. Below is a summary of the results:

ROC Curve

alt text

Key Results:

  • Logistic Regression:
    • AUC: 0.52
    • Performance: The Logistic Regression model's AUC score suggests it is not significantly better than random guessing.
  • Support Vector Machine (SVM):
    • AUC: 0.70
    • Performance: The SVM model performs better than Logistic Regression, with acceptable discriminatory power between the positive and negative classes.

Interpretation:

  • An AUC of 0.5 represents a model with no discriminatory power (random).
  • An AUC of 0.7 indicates acceptable performance, which suggests that SVM is a better model for this task in comparison to Logistic Regression.

Requirements

To run the project, install the necessary dependencies by running:

pip install -r requirements.txt

To run the project after installing the necessary dependencies

python water_potability.py

Employee Turnover Prediction

This project aims to predict whether employees will leave the company using two machine learning models: Logistic Regression and Support Vector Machine (SVM). The dataset is preprocessed, cleaned, and then used to train and evaluate the models.

Table of Contents

  1. Overview
  2. Dataset
  3. Final Accuracy Results
  4. Features
  5. How to Run the Project

Overview

The project involves predicting employee turnover using a dataset with features like:

  • Satisfaction Level
  • Last Evaluation
  • Number of Projects
  • Average Monthly Hours
  • Time at Company
  • Department (encoded)
  • Salary (encoded)

The target variable (left) indicates whether the employee left the company.

Dataset

The dataset is downloaded from the following link:

https://drive.google.com/uc?id=1g1nwk4k-h9FceEHKZc8ocfu_xp3xnZ8R

The dataset contains various employee attributes such as:

  • satisfaction_level
  • last_evaluation
  • number_of_projects
  • average_montly_hours
  • time_spent_at_company
  • Department (Categorical)
  • Salary (Categorical)
  • left (Target variable, 1 if the employee left, 0 otherwise)

Features:

  • satisfaction_level: Employee satisfaction level
  • last_evaluation: Last evaluation score
  • number_of_projects: Number of projects completed
  • average_montly_hours: Monthly working hours
  • time_spent_at_company: Time spent at the company in years
  • Department: Employee's department (Categorical)
  • Salary: Salary category (Categorical)
  • left: Whether the employee left the company (Target)

How to run the project

To run the project, install the necessary dependencies by running:

pip install -r requirements.txt

Run the project after installing the necessary dependencies

python hr.py

Sales Prediction Based on Ad Spend

This is a Streamlit-based web app that predicts sales based on advertising spend in various media channels (TV, Radio, and Newspaper). It supports both single and multiple feature predictions using Linear Regression and Support Vector Regression (SVR) models.

Table of Contents

Overview

This application allows users to predict sales based on advertising expenditure. It accepts a CSV file as input, and the user can select between Linear Regression and SVR models for prediction. The app also supports both single feature (TV ad spend) and multiple feature (TV, Radio, and Newspaper ad spend) predictions.

Features

  • Upload a CSV file to analyze the data.
  • Preview the first few rows of the dataset.
  • Choose between two models:
    • Linear Regression
    • Support Vector Regression (SVR)
  • Choose between:
    • Single feature prediction using only the TV ad spend.
    • Multiple feature prediction using TV, Radio, and Newspaper ad spend.
  • Interactive sliders to adjust ad spend for predictions.
  • Predict sales based on the selected model and feature set.

Usage

  1. Upload a CSV file containing columns like TV, radio, newspaper, and sales.
  2. The app allows you to:
    • Choose a model: Linear Regression or SVR.
    • Choose whether to use a single feature (TV ad spend) or multiple features (TV, Radio, and Newspaper ad spend).
  3. Use the sliders to select the ad spend values for TV, Radio, and Newspaper.
  4. View the predicted sales output.

Input CSV Example

The CSV file should have at least the following columns:

  • TV: TV ad spend (in $)
  • sales: The resulting sales (in millions of dollars)

If you want to use multiple features for prediction, include these additional columns:

  • radio: Radio ad spend (in $)
  • newspaper: Newspaper ad spend (in $)

Example:

TV radio newspaper sales
230.1 37.8 69.2 22.1
44.5 39.3 45.1 10.4
17.2 45.9 69.3 9.3
... ... ... ...

How to Run

To run the project, install the necessary dependencies by running:

pip install -r requirements.txt

After installing the necessary dependencies, run the project

stremlit run ads.py

To run the project in googl colab

!pip install streamlit -q
!pip install streamlit -q

Diabetes Prediction Model

This project applies machine learning techniques to predict whether a patient has diabetes based on various health parameters. Two classification models, Logistic Regression and Support Vector Machine (SVM), are trained and evaluated for their performance on the Pima Indians Diabetes Dataset.

Table of Contents

Overview

This project aims to predict diabetes based on diagnostic measurements from patients. The dataset consists of various health-related parameters like glucose levels, blood pressure, BMI, insulin levels, etc.

Two machine learning models are used for classification:

  • Logistic Regression
  • Support Vector Machine (SVM)

Features

  • Data Preprocessing: Missing values are handled, and certain columns with zeros are replaced by the median value.
  • Exploratory Data Analysis: A correlation heatmap is plotted to identify relationships between features.
  • Modeling: Logistic Regression and SVM are used for predicting diabetes.
  • Evaluation: The performance of the models is evaluated using metrics like accuracy, confusion matrix, and ROC curve.
  • Visualization: Confusion matrices and ROC curves are plotted for both models.

Dataset

The dataset used is the Pima Indians Diabetes Database, which consists of 768 instances and the following features:

  • Pregnancies: Number of times pregnant
  • Glucose: Plasma glucose concentration
  • BloodPressure: Diastolic blood pressure (mm Hg)
  • SkinThickness: Triceps skin fold thickness (mm)
  • Insulin: 2-Hour serum insulin (mu U/ml)
  • BMI: Body mass index (weight in kg/(height in m)^2)
  • DiabetesPedigreeFunction: Diabetes pedigree function (a function that scores likelihood of diabetes based on family history)
  • Age: Age in years
  • Outcome: Class variable (0 or 1), where 1 means the patient has diabetes

Handling Missing Values

Certain columns in the dataset contain zero values, which are not feasible (e.g., 0 for BMI or blood pressure). The columns:

  • Glucose
  • BloodPressure
  • SkinThickness
  • Insulin
  • BMI

are replaced by the median value wherever a zero is encountered.

How to Run

To run the project, install the necessary dependencies by running:

pip install -r requirements.txt

Install the necessary dependencies and run the project

python diabetes.py

To run the project in google colab

1. Install the streamlit

!pip install streamlit -q

2.Check your IP address by running

!wget -q -O - ipv4.icanhazip.com

3. Run the Streamlit app in the background and start localtunnel

!streamlit run ads.py & npx localtunnel --port 8501

It will provide the urls, click on "your url is" it will open the tunnel, enter the ip address generated in step 2 and submit then the app is ready use.