• Predicting future values of stock using Time Series, ML, DL, NLP techniques.
• Time Series Analysis and Forecasting (Moving Avg, Exponential Smoothing, ARIMA, ARCH, GARCH, etc)
• Machine Learning Techniques (Both Classification and Regression)
• Deep Learning Techniques (Simple RNN, GRU, LSTM)
• Prediction using Sentiment Analysis (used VADER's Sentiment Intensity Analyser library)
• UI Building & Deployment using Amazon EC2.
• Various models were compared and following are their respective RMSE scores:

• Genre based Movie Recommendation
• Recommending Movies using Pearson Correlation
• Recommending Movies using Cosine Similarity
• Recommending Movies using KNN (with Cosine distance metric)
• Clustering of movies (using KMeans)
• Recommending Movies using Content & Collaborative latent matrix with TFIDF & SVD
• Predicting unseen Movie ratings for user using Surprise Library (with SVD, KNN, NMF)
• Recommending Movies for a user using Surprise Library (with KNN)

• Created a model that will classify traffic sign images using Deep Learning and Machine Learning techniques.
• Problem has 43 traffic signs (classes) and 50000+ images in total.
• First a Convolutional Neural Network (Convolution >> Batch Normalization >> MaxPool >> Dropout) with fully connected layer (having 2 dense layers) was built which gave an accuracy of 97.11% on test data.
• The same Convolutional Neural Network's output was used to pass onto Random Forest and another model was built. It gave an accuracy of 95.64% on test data.

• Created an app that will predict the ratings of new android apps.
• It will help the android developers to understand the factors behind the rating of any app.
• Data cleaning was done in the first step, followed by extensive EDA.
• Tried to understand univariate and bivariate analysis of the features.
• Analysis was done to understand how each of the categories of the app is performing.
• Performed various hypothesis tests.
• Applied OLS and tried to check autocorrelation and heteroscedasticity.
• Applied Random Forest Regressor, Linear Regression, Support Vector Regressor to select the final model.
• Random Forest Regressor gave best result with RMSE of 0.451.
• Built the UI using HTML, CSS, Flask. Deployment using Amazon EC2.

• Created an app that will predict if a customer is eligible for loan or not.
• Data cleaning was done to fill missing values with mean/median.
• Exploratory Analysis was performed to understand how other features are related to the target.
• Feature engineering was done and new feature Total Income was formulated.
• One Hot Encoding and Label Encoding was performed on Categorical features.
• Relevant features were selected and MinMaxScaler was applied.
• Applied Decision Tree, Random Forest, Gradient Boost, Logistic Regression to select the final model.
• Gradient Boosting model gave best performance with Recall of 53% and F1-Score of 60%.
• Built the UI using HTML, CSS, Flask. Deployment using Amazon EC2.

• Built a binary classification model to detect whether a text message is spam or ham.
• Used NLP techniques like Lemmatization, Bag of Words, TFIDF as a baseline for ML model building.
• Machine Learning models were built using Naive Bayes, KNeighborsClassifier, DecisionTreeClassifier, SVM Classifier.
• For Deep Learning, techniques like Tokenizing, Word Indexing, Sequencing, Padding were followed.
• It was then passed through either the Dense network or LSTM network.
• Among ML models, SVM Classifier using TFIDF gave best result (precision of 100% and F1 score of 92%).
• Among DL models, Dense network performed better in terms of validation loss and accuracy.

ML structure:

DL structure:

• Understanding a retail store's customer purchase data having 51,243 transactions.
• Generating Recency, Frequency, Amount features for RFM analysis.
• Built a 4-segment solution (using K-Means clustering).
• Building segments from the managerial implementation angle.
• Suggesting marketing and promotional activities based on segments.

• Understanding dataset having: Price of Brand X, Feature (value-added variant is present or not), Display (if any promo is present) and its competitor's prices.
• Perform OLS Regression using all features.
• Check significance of features (by P-value) and drop insignificant ones and do OLS again.
• Try feature engineering and check interaction effect between variables and make new features.
• Check adjusted R score and mean absolute error.
• Perform model building using DecisionTreeRegressor, RandomForestRegressor, KNeighborsRegressor.
• Also perform hyperparameter tuning and apply crossvalidation for best result.
• Achieved R2 score of 70% using Random Forest (with interaction effect)

• To predict if a customer is eligible for loan or not.
• Understanding the problem and setting up Spark environment.
• Data cleaning was done and EDA was performed to understand how other features are related to the target.
• Feature engineering was done and new feature Total Income and EMI was formulated.
• String Indexing and One Hot Encoding were performed on Categorical features.
• Using Vector Assembler multiple columns were merged into a vector column.
• Applied Decision Tree Classifier, Random Forest Classifier, Logistic Regression to select the final model.
• Model which gave best performance with high Recall and F1-Score was opted.

• Making a store sales dashboard using Tableau.
• Trying to understand key performing areas and areas which hasn't performed well.
• Understanding how Sales, Profit and Discount varied across Categories, Sub-categories, Regions.
• Graphs used - Scatter plot, Box plot, Butterfly chart, Bump chart, Bar Charts, Maps, Bubbles, etc.