Skip to content

Abhinav-26/Machine-Learning-Minor-Projects

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 

Repository files navigation

Machine-Learning-Minor-Projects

In this repository you can get some minor projects in the Machine Learning domain to learn and practise and increase your Skills. we will be sharing my own small projects which we did during learning ML and understanding the concepts of different algorithms and implementing it.

Machine Learning Fundamentals

Regression
  • Linear Regression:

    • Co2 Emission Prediction
      In this directory you can see a basic machine learning model on Carbon Dioxide Emission Prediction which has been created by using Linear Regression.

  • Logistic Regression

    • Logistic Regression on Churn Dataset
      In this directory you can see a basic prediction using the concept of Logistic Regression. we have used the Churn dataset for predicting customer's churn.

  • Logistic Regression, Decision Tree and K-Nearest-Neighbors

    • Phone Price Prediction
      In this directory we have created a model for predicting the range of price of a phone. The range lies between 1,2,3 (low, medium and high) budgets. We have implemented the concepts of Univariate Feature Selection and Feature Importance for selecting features from the dataset and then created a model for prediction. Also, we have used Logistic Regression, Decision Tree and KNN algorithms and we got better accuracy by using Logistic Regression in this case. We got accuracy around 96% for that.

    • Iris Flower Analysis
      In this ML project , a classification model is made to detect the types of Iris flower. The Iris flower has three classes - Iris setosa, Iris versicolor and Iris virginica.These types of iris flower has different petal lengths & widths, and different sepal lengths & widths. The dataset contains 50 instances of each of these. Logistic Regression, Decision Trees, and K-Nearest Neighbor Algorithms were all used to get accurate and reliable results. Conclusion : One class is linearly separable from the others 2;latter are not linearly separable from each other.

Classification
  • Computer Vision

    • Dog's Breed Prediction Using Transfer Learning
      In this directory we have simply used the pre-trained weights and applied the concept of transfer learning for making our prediction easy. We have used three pre-trained models in this minor project ie, VGG16, Inception_v3 and ResNet50 for predicting the breed of dog. In the File Prediction Comparision I have used the three models for the same image to check which gives the better accuracy and we concluded we got the best accurcay by Inception_v3 model.

    • Cat-Dog Prediction Using CNN
      In this Project I have build a Cat-Dog Prediction model using the concepts of CNN. This is a very basic project of CNN, DeepLearning. This is one of the most beginner friendly project that every beginner usually builds. This project will help you to clear the basic concepts CNN and how it can be implemented in image classification.

    • Eye Filter
      It is an basic eye filter similar to those on instagram and snapchat but it only works with images. Firstly we are detecting the eyes, then reading a filter image which will than replace as eyes. After that we will replace the filter image over the eyes and Congratulations! It's done. For using this project you will need to install python, opencv and numpy and then you need to use a of a person in which eyes are clearly visible. We can also change the color and type of filter as we want but dont play with the fine tuned if statements or it may break the program. Enjoy!

    • Image To PencilSketch.
      We need to read the image in RGB format and then convert it into a grayscale image, or a classic black/white photo. Then the next thing to do is invert the grayscale image (negative image), this will be our inverted grayscale image. This enhances the details. Then we can finally create the pencilsketch by mixing the grayscale image with inverted blurry image. This can be done by dividing the grayscale image by the inverted blurry image. We will need Python's OpenCV library for this.

    • Motion Detector
      The project focused on capturing the video through webcam and then convert each frame to grayscale image or say black&white image we subtract both images from each other. As we subtract only those places will be left as bright place who are diffrent from the previous frame hence motion detected. But at this moment we cant show this to the user and also there will be a lot of noise so we created a threshold value above this all are complete white and below all are complete dark. So after this we need to mark the area in which motion is detected. Therefore we mark that area and then show it to the end user

  • Decision Tree

    • Drug Prescription using Decision Tree
      In this directory we have created a model for predicting the drug which needs to be given to patients having the same disease as per the previous experience. The model created has 100% accuracy, tested on the 30 of the dataset which was not given for training.

  • K-Nearest-Neighbors

    • K-Nearest Neighbors
      In this directory a basic machine learning model is being created for classifying the class of the new customers among 1- Basic Service, 2- E-Service, 3- Plus Service, 4-Total Service.

    • Iris Flower Classification
      It contains the famous Iris Flower Classification project which aims to classify the iris flower into species, Versicolor, Setosa, Virginica.

  • Linear Discriminant Analysis

    • Feature Selection and Modelling on KDDCup Network Intrusion Dataset

      In this project we created a Machine Learning model to check the network is under attack or not. We have used following five feature selection techniques in a sequence to extract the best features from the dataset.

    • Correlation based feature selection

    • Univariate Feature Selection

    • Recursive Feature Elimination (RFE)

    • Principle Component Analysis (PCA)

    • Linear Discriminant Analysis (LDA)

  • Support Vector Machines

    • SVM Cancer Prediction
      In this directory we have created a model using SVM algorithm. The dataset used for creating the model is a record of cell samples of the patients. The dataset consists of 699 records and we got accuracy around 95% with accuracy_score and f1 score.

Clustering
  • Hierarchical Clustering

    • Cars Segmentation using Hierarchical Clustering
      In this directory we have created a car segmentation which may help you to classify the kinds of cars are more popular among customers and may help you to launch your car as per the survey. For this minor project we used the concept Hierarchical Clustering.

  • K-Means Clustering

Advanced Machine Learning

Natural Language Processing
  • Reddit's Flair Detection with Transfer Learning + EDA

    In this directory we have created a Reddits Flair Detector. Here we are not letting the Embedding Layer to initialize it's random weights. We are using a TRANSFER LEARNING approach to train the model. We are using FastText crawl embeddings for transfer learning. These are pre-trained word embeddings trained and released by Facebook after training on 2 million words.

Recommendation System
  • Movie Recommendation

    In this directory we have created a Movie Recommendation system using the concept of content based recommendation system. The dataset used is very large and hence it cannot be directly Uploaded into GitHub. But you can download the data directly from here. Unzip the data and use the respected csv files.

Project Maintainers


Abhinav Dubey


Hritik Gupta