Optimization of Machine Learning Models using Particle Swarm Optimization (An Optimization Heuristic Method based approach) | (self project)
Many Available Machine Learning optimizing tools like grid-search etc optimize a Machine Learning model only to the local minima. Here, I intend to present an Optimization Heuristic Method based approach for optimizing the Parameters of Machine Learning model to the global optima.
It can optimize parameters of any type of Machine Learning / Deep Learning Algorithm.
GitHub Code is available here
Detecting Malariya with Deep Learning using Blood Smears | Kaggle | (self project)
The project's objective was to replace the method of detection of Malariya from manual diagnosis to an automated diagnosis using Deep Learning. Here I have created a CNN from scratch using tensorflow along with the use of some Image processing techniques such as resize, rotate, blur etc. to increase the efficiency of the model over the data-set. The data-set is available
here. The Project code is available
here.
Development of a virtual reality based fire training simulator and machine learning based path guidance system: A case of
hospital fire breakouts | Dr. J. Maiti | Industrial and Systems Engineering | IIT Kharagpur | Duration: 15/05/2019 - 30/06/2019
Created a
virtual environment of a hospital building with a fire in
Unity3D, tested in oculus-rift to collect data from human behavior in case of fire. Implemented a
Reinforcement Learning-based approach on train data from human behavior to bail out of the fire and show the best path. The methodology behind the research can be used for safe evacuation and suggesting new emergency exits from a building in case of fire. Report is available
here and the project certificate is available
here.
Achievements of this project:
Appreciated and encouraged for the research work done by the Ministry of Human Resource Development of India on 3 August 2019.
Accepted for presentation at the 2nd International Conference on Human Interaction & Emerging Technologies: Future Applications (IHIET-AI 2020) to be held at Centre Hospitalier Universitaire Vaudois (CHUV), Lausanne, Switzerland, April 23-25, 2020. (facilitated online on due to COVID-19)
Data Analytics and the Machine Learning model to predict the type of Network Congestion | The Annual Inter-hall Technology General
Championship | Data Analytics Event | Technology Students Gymkhana | IIT Kharagpur
Created a Machine Learning model to predict the type of Network Congestion at a particular area and the particular time using various boosting techniques such as XGBoost optimized the models with grid search and bayesian optimization. Visualized various correlations and features for feature engineering.
Obtained 82.55% train accuracy, 80.66% test accuracy, MCC score of 0.742.
Our Team stood at 1st position based on the MCC score obtained an overall 3rd position out of all participating teams. Awarded with
BRONZE for the Inter-Hall Data-Analytics General Championship organized by The Technology Students Gymkhana, IIT Kharagpur.
The Project Presentation cum report can be found
here . The Team certificate is available
here.
Trends in heart data to predict certain cardiovascular events | (self project)
Data visualization, Interpretation, feature engineering, extracting out some common trend out of the data and plot, Feature modification to obtain some better results out of the model. Please visit
kaggle link to view the project.