leiyunin Goto Github PK

clustering-use-bfr-algorithm

This clustering process iteratively applies K-Means on subdivided data, refining clusters through outlier identification and Mahalanobis Distance measurements, culminating in an optimized segmentation of the dataset into distinct, statistically significant clusters.

data-commons

data-pipeline

decision-trees-as-interpretable-models

This project explores interpretable machine learning models using Decision Trees for classification and LASSO, Ridge Regression, PCR, and Boosting techniques for regression, applied to Acute Inflammations and Communities and Crime datasets, focusing on model interpretability, feature analysis, and optimization.

girvan-newman_detect-communities-in-graphs

This project utilizes the Spark Framework and GraphFrames library to implement the Girvan-Newman algorithm, detecting communities in social networks by analyzing user connections based on common business reviews and optimizing modularity through iterative edge removal.

global-eating-habits-and-diabetes-mortality-an-analytical-study

This project explores the potential relationship between continental eating habits and diabetes mortality rates worldwide. Utilizing data scraped from three web sources, I conducted a detailed analysis that includes visualizing dietary patterns, correlating average caloric intake by continent, and examining diabetes-related death statistics.

hybrid-recommendation-system

This project developed and optimized a hybrid recommendation system that processes over 450,000 training data points and 142,000 validation data points. The system combines user ratings, merchant details, and user reviews to predict users' ratings for restaurants they have not visited.

locality-sensitive-hashing-and-collaborative-filtering-on-yelp-data

The assignment comprises two main tasks: implementing LSH to identify similar businesses based on user ratings and developing various collaborative filtering recommendation systems to predict user ratings for businesses.

mental-health-dashboard

This dashboard is a tool that shows the prevalence and patterns of mental health illness as well as the related health care service based on county-level data among the United States. It also compares the public interests in mental health with the data from Google Search Trends.

multi-class-multi-label-classification-using-svm

This project applies SVM classifiers and K-Means clustering to the Anuran Calls (MFCCs) dataset for multi-class, multi-label classification, evaluating techniques like binary relevance, SMOTE, and Classifier Chains to optimize label prediction accuracy.

realtimereview-flask-websocket

This project creates a real-time review system for LA Veranda Hotel, using Flask and WebSocket to mimic Firebase. It enables easy data manipulation through RESTful APIs and real-time review submissions and monitoring, aiming to improve guest experiences by facilitating instant feedback and management response.

simple-multiple-linear-regression_knn-regression

This project conducts a thorough analysis of the Combined Cycle Power Plant dataset through linear regression, polynomial modeling, and KNN regression, exploring variable interactions and nonlinear associations to predict electrical energy output accurately.

smote-to-a-seriously-imbalanced-dataset

This homework leverages SMOTE for addressing class imbalance in a high-dimensional dataset, employing tree-based methods like random forest and XGBoost with model trees to enhance classification performance on the APS Failure at Scania Trucks dataset.

son-algorithm-implementation-with-spark

This assignment focuses on implementing the SON Algorithm using the Spark Framework to identify frequent itemsets within large datasets.

supervised_semi-supervised_unsupervised-learning

This study evaluates Supervised, Semi-Supervised, and Unsupervised Learning methods on the Breast Cancer Wisconsin and Banknote Authentication datasets, comparing their effectiveness through Monte-Carlo Simulation across multiple performance metrics.

transfer-learning-for-image-classification

This project utilizes transfer learning with pre-trained models EfficientNetB0 and VGG16 in Keras to accurately classify 20 bird species from images, achieving up to 79.53% accuracy with techniques like image augmentation and fine-tuning the last layer.