Insurance Statistical Analysis

This week at ISM, I expanded my knowledge on modelling data and statistics through experiences by taking a dataset from Kaggle and conducting an analysis on it. Following my last visit with Mr. Campbell, I realized the importance of preprocessing data for machine learning modelling prior to feeding it into a reinforcement or deep learning algorithm. Ultimately, the current stage of machine learning, especially at a smaller scale, requires immense help from developers and engineers in order to maximize their effectiveness. Thus, I decided to bolster my skills by taking a sample dataset from Kaggle that covered French automotive vehicle claims and conduct some deeper statistical analysis on them. Last week, I acquired some skills in Jupyter Labs and Notebooks which proved to make the statistical analysis process significantly easier. Ultimately, I was able to integrate my python code, frameworks, and markup code into a single environment with a kernel that made the data presentable. Thus, I was able to compose a "webpage" or notes page that contained all of the data and its visualizations. Based on Mr. Camppell's advice, I was able to construct an array of scatter plots and histograms drawing trends between different factors in the dataset for insurance claims. After that, I was able to put all of the datasets together into a heatmap that drawed connections between all of the correlations we made to create a larger analysis that spans over all of the attributes. After hours of processing through hundreds of thousands of samples, I was ultimately able to draw some faint trends in the data. This experience really revealed the importance and scale of big data in machine learning. It required hundreds of thousands of samples from organized datasets to draw simple trends in insurance claims. It makes me wonder how much data must be collected, organized, and processed in a Google search algorithm or recommendation system. This experience over the last week really put machine learning into perspective for me and I will have new lesions going forward as I continue the modelling stage of my final product.