Student AI / ML Projects with Python
Student AI / ML Projects with Python
Embedded Files
You can access our application DYL-HP here.
This project aims to segment customers of an omnichannel retail company (FLO) using unsupervised machine learning techniques. Understanding customer groups is essential for tailoring marketing strategies, improving customer experience, and increasing long-term profitability.
Customer segmentation will be achieved by analyzing behavioral data such as purchase frequency, transaction value, recency, and customer tenure. The goal is to group customers with similar patterns and create targeted strategies for each group.
This study develops a weekly purchase-order forecasting system for a textile firm using SARIMA. After log-transforming and ensuring stationarity via ADF tests, p,d,q and P,D,Q parameters were tuned by AIC-based grid search. The chosen SARIMA(1,0,1)×(0,1,0,52) model achieved over 99 % explained variance and strong MAE/RMSE performance. The methodology—combining statistical rigor, hyperparameter tuning, and outlier treatment—is reproducible and scalable to other seasonal forecasting problems.
This guide is based on a complete machine learning project aimed at forecasting product demand in the electronics sector using real-world sales data from Kaggle. The system uses regression models within a modular software architecture, enabling automation, scalability, and high predictive accuracy.
This document details the development of an automated system designed to visually detect defects in solar panels on a production line. The system leverages image processing techniques and machine learning to identify faulty panels, thereby improving quality control and production efficiency. The modular design, comprising data loading, feature extraction, model training, and evaluation components, allows for flexibility and scalability.
In today’s world, industries are moving faster than ever. Factories are becoming more automated, and quality matters a lot, especially when it comes to materials like concrete. Even a tiny crack in a concrete surface can lead to big problems later, especially in things like bridges, buildings, or roads. Usually, human workers check these concrete surfaces by hand. But this kind of manual inspection takes a lot of time and isn’t always accurate. People get tired, miss small defects, or might just have an off day. So, we decided to build a smarter solution—a computer system that can look at photos of concrete and instantly tell whether there’s a defect or not. It’s like having a digital inspector that never gets tired, never blinks, and always works at full speed. The system we built uses something called machine learning. This is a way of teaching a computer how to recognize patterns by showing it lots and lots of examples. In our case, we showed it thousands of images of concrete, some with cracks, and some without. Over time, the system learned what “good” and “bad” concrete looks like, just like a human would if they were shown enough examples. Our goal was to create a system that could:
• Help spot defects earlier in the production process
• Save time by speeding up inspections
• Improve quality control by reducing human error
And most importantly, we wanted it to be accurate, fast, and easy to use in real factories.
This guide summarizes a machine learning project aimed at predicting customer churn using bank customer data. The system implements classification algorithms within a modular software architecture to achieve automation, flexibility, and high accuracy
In this project, a real-time sound classification and processing system powered by artificial intelligence was developed to improve communication in noisy industrial environments. The system classifies audio signals into three categories—Human, Machine, and Mixed—using supervised learning algorithms and applies corresponding signal processing techniques to enhance intelligibility. Feature extraction was performed using Mel-Frequency Cepstral Coefficients (MFCC), and three machine learning models (Random Forest, SVM, and KNN) were tested. Among them, K-Nearest Neighbors (KNN) was selected due to its superior performance in identifying mixed audio. In the case of mixed signals, Demucs, a deep learning-based source separation model, was utilized to isolate the human voice. The entire system was designed to operate efficiently even for users with no programming knowledge. Results show that the system effectively enhances communication by amplifying human speech and suppressing background machinery noise in real-time.
The primary goal of this project is to optimize traffic light timing and enhance traffic flow efficiency. The aim is to minimize traffic congestion, especially in areas with heavy traffic, and reduce vehicle waiting times. To achieve this, a decision tree model was developed by analyzing historical traffic data.
This project aims to predict the market capitalization of companies across different sectors using historical financial data and stock prices.To ensure a clear structure, I organized the project into separate folders for data, business logic, utilities, and results. This guide provides a step-by-step walkthrough of each component in an accessible and well-organized manner
The primary goal of this project is to predict cotton lint yield based on agricultural input features using machine learning techniques. Accurate yield prediction models are essential in agriculture for optimizing resources and improving productivity. Therefore, this project aims to develop an explainable and high-accuracy predictive model to support data-driven decision-making in cotton farming.
Google Sites
Report abuse