Machine-Learning

Graduate coursework in supervised and unsupervised learning, model evaluation, feature engineering, and deep learning for classification and regression, with business and computer-vision applications.

---

1. Title and Summary

Machine Learning
Northwestern University M.S. in Data Science (Data Engineering specialization): end-to-end ML workflow from data prep and baseline models through tree ensembles, dimensionality reduction, clustering, and TensorFlow/Keras neural networks on tabular credit-risk data, steam benchmarks, iris, Fashion-MNIST, and satellite/cat-dog imagery.

---

2. Concepts and Methods

• ML framework and tabular EDA: load and inspect loan default datasets (HMEQ_Loss.csv); target variables for bad-flag classification and loss amount regression; missing-value and feature exploration (Machine Learning.ipynb)
• Linear and logistic regression: OLS on steam heating data with R² and MSE; logistic regression on steam logit data (Linear_Logistic_Regression.ipynb)
• Feature scaling / transformation: MinMaxScaler and StandardScaler on HMEQ features; compare normalized vs. standardized feature frames (Transform.ipynb, Transform.py)
• Regression model selection: stepwise and sequential feature selection with mlxtend SequentialFeatureSelector; compare linear, tree, forest, boosting, and Keras dense models; ROC/AUC curves for classifiers (Regression_Models_Stepwise_Sequential_Feature_Selection.ipynb)
• Tree-based methods: decision trees, random forests, gradient boosting for classification and regression; confusion matrices and classification reports (DecisionTrees_RandomForests_GradientBoosting_Receiver_Operator_Characteristic_Curve.ipynb)
• Unsupervised learning: K-means on iris and HMEQ with silhouette and Calinski-Harabasz scores; elbow-style cluster evaluation (KMeans Clustering.ipynb, K_Means_Clustering.py, KMeans PCA.ipynb, K_Means_PCA.py)
• Dimensionality reduction: PCA scree plots, variance explained, component loadings, 2D projection by species (Principle Component Analysis.ipynb, Principal_Component_Analysis.py)
• Fully connected neural networks (tabular): Keras Sequential dense/dropout models on HMEQ for binary default classification and loss regression; architecture and activation sweeps (Deep Learning with Artificial Neural Networks.ipynb, overlap in stepwise/ROC notebook)
• Fashion-MNIST MLP/CNN baselines: load fashion_mnist; flatten/normalize pixels; dense and convolutional experiments (Neural_Networks.ipynb, Neural_Network_MNIST_Fashion_Data.ipynb)
• Computer vision pipeline: OpenCV grayscale load/resize; cat/dog and satellite image folders; pickle feature tensors (X.pickle, Y.pickle); pixel normalization comparisons (Data Normalization for Image Classification.ipynb, Satellite Imagery Machine Learning Image Recognition Pixel Normalization.ipynb)
• Convolutional neural networks: Conv2D + MaxPooling stacks vs. dense baselines on pickled image tensors; train/validation split; softmax binary classification (Deep Learning with Convolutional Neural Networks.ipynb)
• Model inference on new images: load saved Keras model (TFNN.pet.model); preprocess and score held-out pet imagery (Applying CNN to New Data.ipynb)

Data dependencies: coursework CSVs (HMEQ_Loss.csv, Steam_Linear_Data.csv, Steam_Logit_Data.csv, IRIS.csv), image directories, pickle tensors, and saved Keras models are referenced in notebooks but not bundled in the repository

---

3. Stack

Layer	Tools
Language	Python 3
Environment	Jupyter Notebook
Classical ML	scikit-learn (linear/logistic regression, trees, forests, gradient boosting, KMeans, PCA, metrics)
Feature selection	mlxtend (SequentialFeatureSelector)
Deep learning	TensorFlow 2 / Keras
Vision	OpenCV (cv2), pickle serialization
Data / viz	pandas, NumPy, matplotlib, seaborn
Scripts	Transform.py, K_Means_Clustering.py, K_Means_PCA.py, Principal_Component_Analysis.py

---

4. Structure

Machine-Learning/
├── Machine Learning.ipynb
├── Linear_Logistic_Regression.ipynb
├── Transform.ipynb
├── Regression_Models_Stepwise_Sequential_Feature_Selection.ipynb
├── DecisionTrees_RandomForests_GradientBoosting_Receiver_Operator_Characteristic_Curve.ipynb
├── KMeans Clustering.ipynb
├── KMeans PCA.ipynb
├── Principle Component Analysis.ipynb
├── Deep Learning with Artificial Neural Networks.ipynb
├── Neural_Networks.ipynb
├── Neural_Network_MNIST_Fashion_Data.ipynb
├── Data Normalization for Image Classification.ipynb
├── Satellite Imagery Machine Learning Image Recognition Pixel Normalization.ipynb
├── Deep Learning with Convolutional Neural Networks.ipynb
├── Applying CNN to New Data.ipynb
├── Transform.py
├── K_Means_Clustering.py
├── K_Means_PCA.py
├── Principal_Component_Analysis.py
└── README.md

• Organization: notebooks grouped by method (tabular supervised → unsupervised → deep learning tabular → vision); four standalone .py scripts mirror notebook exercises
• Reusable modules: none packaged; helper functions (e.g., ROC plotting, image loaders) defined inline per assignment
• Engineering practice: train/test splits, ROC/AUC model comparison, sequential feature selection, cluster validation metrics, scree-driven PCA truncation, CV-based image preprocessing pipelines, saved-model inference on new files

---

Course context: Northwestern University, M.S. in Data Science, Data Engineering specialization
Repository: https://github.com/EAName/Machine-Learning