core-datascience-models

A structured collection of machine learning notebooks covering core algorithms from supervised and unsupervised learning to NLP and time series forecasting. Each notebook includes line-by-line comments explaining both the code and the underlying concepts, with references where applicable.

---

Repository Structure

core-datascience-models/
├── Supervised_Learning/
│   ├── DecisionTree_CART_Classification.ipynb
│   ├── NaiveBayes_FromScratch_SpamDetection.ipynb
│   ├── NaiveBayes_TestScript_Iris.ipynb
│   ├── SVM_AirQuality_Pollution.ipynb
│   ├── SVM_Classification_Iris.ipynb
│   └── SVM_HeartAttack.ipynb
├── Unsupervised_Learning/
│   └── Clustering_AssociationRuleMining.ipynb
├── Regression_and_Neural_Networks/
│   ├── LogisticRegression_PoissonRegression_AdultIncome.ipynb
│   └── NeuralNetwork_Keras_LoanPrediction.ipynb
├── NLP/
│   └── SentimentAnalysis_NLP_Tweets.ipynb
├── Time_Series/
│   └── TimeSeries_ARIMA_Forecasting.ipynb
└── data_mining_practice/
    ├── Chapter_01_Data_Loading.ipynb
    ├── Chapter_03_Data_Preparation.ipynb
    ├── Chapter_04_EDA.ipynb
    ├── Chapter_05.ipynb ... Chapter_13.ipynb
    ├── SA01_BankMarketing.ipynb
    └── SA02_ChurnPrediction.ipynb

---

Notebooks

Supervised Learning

Notebook	Algorithm	Dataset	Core Concept
DecisionTree_CART_Classification	Decision Tree (CART)	Custom	Recursive splitting, information gain, Gini impurity
NaiveBayes_FromScratch_SpamDetection	Naive Bayes	Spambase (UCI)	Built from scratch without sklearn — Gaussian likelihood, prior/posterior calculation
NaiveBayes_TestScript_Iris	Naive Bayes	Iris	Manual accuracy function, model evaluation
SVM_Classification_Iris	SVM	Iris	Hyperplane optimization, kernel tricks
SVM_AirQuality_Pollution	SVM	Air Quality / Pollution	Seaborn pair plots, margin analysis on real-world data
SVM_HeartAttack	SVM	Heart Attack Risk (PH)	Feature exploration, classification on health data

Unsupervised Learning

Notebook	Algorithm	Dataset	Core Concept
Clustering_AssociationRuleMining	K-Means, Association Rules	Custom	Centroid clustering, Apriori, support/confidence/lift

Regression and Neural Networks

Notebook	Algorithm	Dataset	Core Concept
LogisticRegression_PoissonRegression_AdultIncome	Logistic Regression, Poisson Regression	Adult Income (UCI)	MLE via statsmodels, multicollinearity check, model comparison
NeuralNetwork_Keras_LoanPrediction	Feedforward Neural Network	Loans (49,698 records)	Keras Sequential API, ReLU + Sigmoid activations, forward propagation

NLP

Notebook	Algorithm	Dataset	Core Concept
SentimentAnalysis_NLP_Tweets	Sentiment Analysis	Kaggle Tweet Dataset	Text preprocessing, emoticon-based labeling, binary classification

Time Series

Notebook	Algorithm	Dataset	Core Concept
TimeSeries_ARIMA_Forecasting	ARIMA	Custom	pmdarima auto-ARIMA, stationarity testing, forecast evaluation

---

Data Mining Practice (data_mining_practice/)

Textbook exercises following a structured data mining workflow — from data loading and preparation through EDA, modeling, and evaluation. Uses the Bank Marketing and Churn datasets from the UCI Machine Learning Repository.

---

Tools and Libraries

• Python 3.11
• scikit-learn — model training and evaluation
• statsmodels — logistic and Poisson regression
• keras / tensorflow — neural network construction
• pandas, numpy — data manipulation
• matplotlib, seaborn — visualization
• pmdarima — auto-ARIMA for time series

---

Notes

• All notebooks were developed and run on Google Colab.
• The NaiveBayes_FromScratch notebook implements the algorithm without sklearn to demonstrate understanding of the underlying math.
• Datasets used are either publicly available (UCI ML Repository, Kaggle) or course-provided.