Month: November 2023

Today, our group got together to talk about our second project, which looks at data concerning police shootings in Washington. We reviewed our progress and the questions we’ve been exploring. I’ve been using a statistical technique called logistic regression, tweaking it to get clearer insights. I plan to use more techniques to dig deeper into the data. Our team is keen on uncovering helpful insights from these numbers, We’ll keep trying different approaches to really understand these critical incidents.

We have closely examined the distribution of police shooting incidents in clusters around the United States. We discovered some places with higher numbers of these episodes and determined where they occur more frequently using two distinct methods: K-Means and DBSCAN. DBSCAN identified particular regions and odd patterns, whereas K-Means provided us with a clear division of the instances. Combined, these techniques improved our understanding of the geographic locations and patterns of these occurrences, which is a useful place to start for anyone wishing to learn more about this problem or develop solutions.

I completed a thorough Linear Discriminant Analysis (LDA) on the provided dataset concerning fatal police shootings.Linear Discriminant Analysis (LDA) on our recent dataset involving fatal police shootings. The LDA process was methodically carried out, beginning with a careful selection of features that influence the outcome of such incidents. We preprocessed the data, normalizing and scaling where necessary to ensure uniformity and fairness in analysis. The core of LDA involved calculating eigenvalues and eigenvectors to determine the linear combinations that best discriminate between the classes. We then projected the data onto a new axis to maximize the separation between various groups. The visualization of the results provided us with a stark representation of the distinct categories within the data, reinforcing the discriminative power of LDA in pattern recognition and classification.

I have completed a clustering analysis using the DBSCAN algorithm, a powerful technique that groups data into clusters based on their similarity. For the dataset, which contains information on fatal police shootings, we focused on clustering incidents by age, and geographical location (longitude and latitude). This method helped us identify natural groupings within the data without the need to pre-specify the number of clusters. Our initial results have revealed 5 distinct clusters, providing us with insights into patterns that may exist within these tragic events. The clusters represent data points that are densely grouped together, while points that did not fit into any cluster were marked as outliers. This is a step forward in understanding the underlying structure of the data, which could be pivotal for further analysis.