Akansha Singh

Postgraduate in Data Science & Business Analytics from IIT Kanpur with 6+ years of professional experience in software development. Seeking to transition into data science, leveraging strong analytical training, problem-solving skills, and a passion for contributing to the data science community through knowledge sharing and impactful projects.

Emotion Detection using Facial Recognition  2017- 2018
A final year project on machine learning to identify human emotion. Using facial expressions using python with tensor flow.
https://www.ijcseonline.org/pub_paper/19-IJCSE-03027.pdf

Machine Learning Projects : https://github.com/singh2akansha/Machine-Learning-Projects

Final Year Project – Time Series Forecasting of Copra & Coconut Oil Prices 

• Built an end-to-end time series forecasting pipeline using Prophet, XGBoost, and Ensemble (Prophet+XGBoost) to predict coconut oil prices with 99.99% R² and 0.10% MAPE, outperforming ARIMAX/VARX baselines.
• Engineered domain-specific features including monsoon indices, MSP revision flags, festival demand effects, lagged price differentials, and global commodity trend signals to improve model accuracy and interpretability.
• Conducted in-depth trend decomposition, changepoint detection, ACF/PACF, and residual diagnostics across 311 months of historical price data.
• Designed benchmarking framework comparing statistical vs machine learning forecasting models, demonstrating 15– 20% improvement in predictive precision using hybrid models.
• Delivered a scalable forecasting solution for procurement and policy decision-making in the agri-commodity sector.

Supervised Learning 

Airbnb Stock Price Prediction

• Built regression models (Decision Tree Regression, Random Forest) to forecast Airbnb stock prices.
• Performed feature engineering, scaling, and model tuning to improve RMSE and predictive accuracy.

Airlines Delay Prediction

• Predicted flight delays using supervised classification algorithms (Decision Tree, Random Forest).
• Analyzed weather, schedule, and route-based features to identify key delay drivers.

Diabetes Prediction using Classification and Boosting Algorithm

• Developed models using Decision Tree Classification, Random Forest Classification, AdaBoost and XGBoost to predict diabetes likelihood.
• Applied feature scaling, missing value imputation, and evaluated models with accuracy, precision, and recall metrics.

Unsupervised Learning 

Customer Segmentation Using Clustering Techniques

• Implemented K-Means and Hierarchical Clustering for customer segmentation.

Fraud Transaction Detection using Anomaly Detection

• Built an unsupervised fraud detection model leveraging both Isolation Forest and Local Outlier Factor (LOF) algorithms to detect anomalies in financial transactions.

Market Basket Analysis

• Implemented Apriori Algorithm to identify frequent itemsets in transactional datasets based on minimum support thresholds.
• Performed Association Rule Mining to generate rules from frequent itemsets and calculate support, confidence, and lift for evaluating the strength of product associations.
• Derived actionable insights for cross-selling and recommendation systems in retail environments.

Exploratory Data Analysis : https://github.com/singh2akansha/Exploratory-Data-Analysis

Diabetes Prediction Using Classification Algorithm (EDA)

• Performed data cleaning, visualization, and statistical exploration to identify key factors influencing diabetes.
• Used pair plots, heatmaps, and correlation analysis for data insights before model development.

Flight Delay Prediction (EDA)

• Conducted end-to-end exploratory analysis on flight delay data.
• Visualized delay patterns by airline, time of day, and route using Seaborn and Matplotlib to derive operational insights.