Souryadip's Projects
In this project I analyzed some of the major cryptocurrencies and also forecasted the price of BTC using LSTM model. The result of the forecast didn't come out too well. Time series forecasting is a challenging thing to do and there are many exogenous variables that can influence the time series data which justifies the output of the forecasting.
In this project I scrapped data from Sulekha.com about Nurseries in Delhi NCR region and made a dataset from the scraped data, then cleaned the datset using python code and also manually using excel and added few more attributes manually with excel such as Website link and Social media link of the nurseries by personally finding them by searching with the nursery name on google. Then, I made an graphical analysis of the data using python and generated a report.
The aim is classify 11 different types of fashion apparels. The train set contains 31719 images and the classes are imbalanced. I have used the SMOTE algorithm to handle the imbalanced class problem then I trained one of the state of the art transfer learning models with some tweaks to achieve training accuracy of about 99% and validation accuracy of about 87% after 21 epochs of training. I have written this code for an ongoing kaggle competition so I am refraining myself from sharing the model architecture.
I have created a database and imported three tables into the database. Then, I ran several SQL queries to find out several facts related to IPL matches.
The notebook provides our solution to Hackathon 2022 hosted by IIT Madras. We achieved a rank of 22 out 69 participants.
I have written a python class that you can use to generate bootstrap samples or Monte Carlo simulation samples and also calculate the confidence interval of mean for a given confidence level.
prompt2model - Generate Deployable Models from Natural Language Instructions
In this project I have created a sentence similarity score calculator which takes two sentences as input and shows their corresponding similarity score.
Config files for my GitHub profile.
I predicted the survival of passengers using the famous Titanic Dataset. I used Decision Tree Classifier model for the prediction.