🏥 Patient Satisfaction Prediction using Machine Learning

Patient Satisfaction Prediction is a machine learning project that predicts patient satisfaction levels using doctor related metrics. Designed with efficiency, clarity, and model benchmarking in mind, complete with automated evaluation.

---

📊 Project Overview

This project explores how doctor experience, fees, wait times, and other metrics impact patient satisfaction. It applies various machine learning techniques to:

• 📌 Preprocess and normalize raw data
• 🤖 Train and evaluate multiple models (manually + via LazyPredict)
• 📈 Benchmark performances
• 📁 Save results for further analysis

---

🧠 Features

• ✅ Handles missing values automatically
• 📐 Normalizes numerical data using MinMaxScaler
• ⚖️ Balances dataset using SMOTE
• 🧪 Trains multiple classifiers including:
  • Logistic Regression
  • Random Forest
  • Decision Tree
  • AdaBoost
  • KNN
  • SVM
  • Naive Bayes
  • XGBoost
• 🤖 Uses LazyPredict to automatically benchmark many models
• 📊 Saves model performance comparisons as CSV
• 🎯 Voting Classifier for accuracy boosting
• 📦 Cleanly modular and well-commented code

---

🗂️ Dataset

• Location: ./dataset/doctors_dataset.csv
• Contains:
  • Experience (Years)
  • Total Reviews
  • Satisfaction Rate (%)
  • Avg Time to Patients
  • Wait Time
  • Fee (PKR)
  • Target label (Satisfaction: Satisfied / Unsatisfied)

---

🧪 Algorithms & Tools Used

Type	Tools/Libraries
Programming	Python 3.11+
ML Frameworks	scikit-learn, xgboost, lazypredict, imblearn
Visualization	seaborn, matplotlib
Data	pandas, numpy
IDE	Jupyter Notebook / VS Code

---

🔄 How to Run the Project

1. Set up the environment

# Clone the repo
git clone https://github.com/Usama-Codez/Patient-Satisfaction-Prediction.git
cd Patient-Satisfaction-Prediction

Create & activate a virtual environment

python -m venv newenv
newenv\Scripts\activate

Install dependencies

pip install -r requirements.txt

2. 🧠 Run the notebook

You can Use any of the following IDEs:

• Jupyter Notebook
• Jupyter Lab
• Google Colab Make sure to select the correct environment when running your notebook.

📉 📈 Confusion Matrix, ROC curve and Evaluations:

Correlation Heat Map:

Confusion Matrix Logistic Regression model:

ROC Curve Logistic Regression model:

Evaluating LogisticRegression

Accuracy: 58.76% Precision: 0.95 Recall: 0.59 F1 Score: 0.73 ROC-AUC: 0.5667822806377023

Confusion Matrix Random Forest Classifier:

ROC Curve Random Forest Classifier:

Evaluating Random Forest Classifier

Accuracy: 90.58% Precision: 0.95 Recall: 0.95 F1 Score: 0.95 ROC-AUC: 0.735426554703663

Confusion Matrix XGBoost:

ROC Curve XGBoost:

Evaluating XGBoost

Accuracy: 90.96% Precision: 0.94 Recall: 0.96 F1 Score: 0.95 ROC-AUC: 0.7725142996227334

Confusion Matrix SVM:

ROC Curve SVM:

Evaluating SVM

Accuracy: 72.32% Precision: 0.95 Recall: 0.75 F1 Score: 0.84 ROC-AUC: 0.6031702567847147

Confusion Matrix KNN:

ROC Curve KNN:

Evaluating KNN

Accuracy: 76.27% Precision: 0.94 Recall: 0.80 F1 Score: 0.86 ROC-AUC: 0.5284166970913959

Confusion Matrix Decision Tree:

ROC Curve Decision Tree:

Evaluating DecisionTree

Accuracy: 79.66% Precision: 0.95 Recall: 0.83 F1 Score: 0.88 ROC-AUC: 0.6035049288061336

Confusion Matrix NaiveBayes:

ROC Curve NaiveBayes:

Evaluating NaiveBayes

Accuracy: 63.09% Precision: 0.95 Recall: 0.64 F1 Score: 0.76 ROC-AUC: 0.5259827187538031

Confusion Matrix AdaBoost:

ROC Curve AdaBoost:

Evaluating AdaBoost

Accuracy: 77.21% Precision: 0.96 Recall: 0.79 F1 Score: 0.87 ROC-AUC: 0.7437325057806986

Testset evaluations:

Fitting 3 folds for each of 108 candidates, totalling 324 fits Best Parameters: {'max_depth': None, 'min_samples_leaf': 1, 'min_samples_split': 2, 'n_estimators': 100} Best Cross-Validation Accuracy: 0.9363507779349364

Author: Usama Akram

📌 License

This project is licensed under the MIT License.