Application of Swarm Intelligence to Portfolio Optimization

Portfolio Optimization, Swarm Intelligence, Microbat Echolocation, Elephant Herd Optimization Akhil Sethia Sept'18
Compared the feasibility, convergence and performance of Cuckoo Search, Firefly Algorithm, Microbat Echolocation, Elephant Herd, Flower Pollination Algorithm, Harmony Search, Differential Evolution & Particle Swarm Optimization for the portfolio optimisation problem. Used Sharpe Ratio and 99% Value-at-Risk weighted return as the objective function.
Presented a technical paper for the same in IEEE's International Conference on Computing, Power and Communication Technologies(GUCON 2018) with Record No. 44295 and has been submitted to IEEE's Digital Xplore for further publishing.

Data Augmentation using Generative models for Credit Card Fraud Detection

Generative Adversarial Networks, Credit Card fraud, Anomaly detection Akhil Sethia, Raj Patel & Purva Raut Dec'18
Generated dummy credit card fraud data using Generative Adversarial Networks to solve the class imbalance problem in the dataset. Used an artificial neural network for detecting fraudulent records and generated dummy data using vanilla, Wasserstein, Relaxed Wasserstein, Least Squares and Margin Adaptive GAN's. Compared the distribution of data produced, training convergence of models and the relative performance of the architectures.
Presented a technical paper for the same in IEEE's 4th International Conference on Computing Communication and Automation (ICCCA 2018) with Record No. 4295XP and has been submitted to IEEE's Digital Xplore for further publishing.

Stock Price Prediction using LSTM, GRU & ICA

Long Short Term Memory, Gated Recurrent Unit, Time Series Forecasting Akhil Sethia & Purva Raut April'18
Predicted the t+5th day adjusted close price for an equity script. Used a LSTM and GRU based neural networks to make these predictions. 45 technical analysis based attributes were computed from tick data, ICA was then applied to it to extract the 12 most relevant features. The inputset was then fed in to a multi-layered recurrent neural network and the performance was compared with other state of the art algorithms analysed using self-designed metrics.
Presented a technical paper on the same in the Springer ICTIS 2018 which is published as the 46th chapter in Information and Communication Technology for Intelligent Systems, Volume 2 book having a DOI of https://doi.org/10.1007/978-981-13-1747-7_46.

Time Dependent Analysis of Machine Learning Algorithms

Supervised Learning, Time based Evaluation, Performance Comparison Akhil Sethia Feb'18
Developed a technique or scaling the performance of machine learning algorithms based on their training and testing times and used it to analyze 10 supervised learning classifiers based on their average performance over 10 imbalanced and balanced datasets. Analysed the optimal use-case of each algorithm dependent upon its time based training and testing efficiency and its ability to handle imbalanced datasets.
Published a technical paper on the same in IJSR having an ISSN 2319-7064 and DOI: 10.21275/ART2018335.

Blockchain - Future of Decentralised Systems

Blockchain,Crypto-currency, Consensus, Smart Contracts Akhil Sethia, Raj Patel & Shyam Patil Sept'18
Co-researched on the working mechanism, security, vulnerabilities, scope and applications of blockchain technology. Researched on its applications to smart contracts, cloud based storage, crypto-currencies, steem and other use-cases.
Presented a technical paper for the same in IEEE's International Conference on Computing, Power and Communication Technologies(GUCON 2018) with Record No. 44295 and has been submitted to IEEE's Digital Xplore for further publishing.