Time series anomaly detection and RAG system for AI-driven governance insights projects by Aspire to Innovate (a2i)

Abstract

In the field of Machine learning, there are many applications where we can add our trained model to use. One of the most useful this is using LLM models but often these LLM models can not give us the latest answer or answer based on our needs. This is where RAG comes in this gives LLM models context. It works like a helping hand to the LLM models giving it proper context so that it performs based on our needs. Many advancements in machine learning and artificial intelligence have occurred in recent years. To make data-driven decisions, we need a lot of data. Collecting data is a big part of Machine learning and data science. The large language model needs a lot of data to make meaningful answers. Web scraping is a big part of that. Finding data for any Bangla language can be hard as there are not many sources that provide it. For this, we are highly dependent on web scraping. Our project- specific data will be gathered by scrapping the popular Bangla news portal website. This project aims to scrap a news portal website from a given date range. The user can select the range of the data and which news portal they want to scrap from. Time series analysis and prediction is also one of the crucial aspects of machine learning oftentimes time it does not follow traditional norms of prediction models like regression and classification. Throughout my internship, I have worked with time series prediction models which helps businesses to predict the future and be prepared. Throughout my internship period, I was involved in various projects like time series analysis and prediction, sentiment analysis, creating a scrapping website, and mak- ing RAG with LLM models. This report contains all the necessary information and processes of those projects which shows my journey as a machine learning intern in a2i