While conducting my research at IISER Kolkata, I wanted to take a few moments to share an exciting project I have been working on, which I believe holds significant potential for real-world impact.

As we know, PageRank, the renowned algorithm originally developed by Google, ranks web pages based on the number and quality of incoming links. I explored an innovative application of this algorithm in the stock market domain, which is inherently complex and driven largely by mass psychology and the influence of a relatively small group of dominant market players.

Research Approach:

• Market Structure Assumption:
  I began with the understanding that stock prices are significantly influenced by the collective psychology of the market, where a few successful traders, institutional investors, and influential entities largely dictate trends. The majority of individual investors tend to follow these leaders, often creating cascading effects in the market.

• Trader Classification:
  I systematically grouped successful and unsuccessful traders based on their historical performance and behavioral patterns. By mapping their investment decisions, I constructed an investment influence graph analogous to the hyperlink structure in web networks.

• PageRank Application:
  I applied the PageRank algorithm to this trader influence graph to quantitatively rank companies based on their sensitivity and importance within the network of influential traders. This produced a prioritized list of companies whose stock prices are more susceptible to influential trading activities.

• Dynamic Graph Modeling:
  To adapt to the fast-changing nature of financial markets, I integrated dynamic graph techniques that update the influence network in near real-time, ensuring the model remains responsive to evolving market conditions.

• Predictive Modeling:
  Building on the PageRank-based insights, I developed a predictive model using a combination of Genetic Algorithms and Machine Learning, including Neural Networks. This hybrid approach helped optimize the feature selection, learning process, and prediction accuracy.

Key Contribution:

The outcome of this research is a dynamic, adaptive system for stock price prediction that not only considers traditional financial indicators but also integrates behavioral and network-based factors. I believe this framework could significantly enhance our understanding of market movements and contribute to more robust trading strategies, benefiting both individual investors and financial institutions.