Projects

Taxi Fare Prediction | Prof. Faiz Ahmed

  • Designed model which predicts fare on various factors, with optimal accuracy.
  • Performed EDA on data using data visualization tools and techniques.
  • Employed Linear regression, Lasso, Random Forest and Artificial Neural Network models on the pre-processed data
  • Model predicted fare with error ±5% compared to actual fare predictor device

Fake News Detection | Prof. Arnab Bhattacharya

  • Designed model to classify articles as real or fake and act as a filter for articles
  • Performed pre-processing, tokenization, word stemming to convert text documents into features vector using NLTK
  • Used feature set as a body text, title, body + title, and applied Logit, Naïve Bayes, ANN, and Support vector machine for classification
  • Compared results obtained from different classification technique, out of which SVM gave an average accuracy of 93%.

House Price Prediction Using MLR | Prof. Sharmishtha Mitra

  • Developed a Multiple Linear Regression model using Ordinary least square (OLS) method to predict ‘House price’, using data from Kaggle.
  • Explore data using statistical technique, checked problem of multicollinearity, detect outlier, checked the significance of the regressor, checked heteroskedasticity, checked leverage and influential points, checked model adequecy, did residual analysis, selected good features using Backward elimination

Rule Mining Using ANN | Prof. Arnab Bhattacharya

  • Rule mining using ANN, taking Real world data and then figuring out the rules that govern the data.
  • Implemented this method to obtain rules from ANN on training data for binary classification. Employed mini-batch training with Adam optimizer to obtain competitive performance.
  • Used different approaches like SVM, Decision forest etc. to serve as a baseline for comparing performance on the dataset.

Compartment Model | Prof. Debasis Kundu

  • Estimated the parameters of Compartment model using Prony’s equations, Newton Raphson or Gauss Newton method and Osborne’s method.
  • Plotted the curve between actual verses predicted to see the error and Computed the confidence intervals of the unknown parameters.