Investigating the Socioeconomic Dynamics of Crime: Exploring the Relationship Between Income and Crime Rates in Canada

This project investigates the correlation between crime rates and income levels across Canadian provinces by merging and analyzing datasets from Statistics Canada.
The study reveals a general trend where higher average incomes are associated with lower crime rates, although the strength and nature of this relationship vary by region. Key insights include significant regional differences in crime types and rates, with notable socioeconomic implications.

Data Analysis, R (Programming Language), Statistical Modeling

Rising concerns label foreign interference as a critical threat to Canada's national security, notably via social media. Recognizing its global influence, we scrutinized 758 Chinese social media entities across six platforms—X (Twitter), Facebook, Instagram, Youtube, Threads, and TikTok.
Key aspects include scrutinizing the top 12 Twitter accounts, evaluating CGTN Accounts versus Chinese Embassy Accounts, and presenting our insights visually through slides. Despite dataset limitations..

Data Visualization, Data Analysis, Python, Matplotlib, NumPy Pandas

Developed an food ordering system targeting on food trucks in the campus and student consumers, to reduce wait time by efficient food pick-up;
Built use-ends for both food retailers and consumers, equipped with Action Listeners to achieve real-time communication and event handling;
Utilized MongoDB to store and manage user data instead of CSV file, reduced complexity of data management and query by 50%.

Java, JavaSwing, AWT, MongoDB, Javax.mail, Git

Github Repo

Conducted a research on the impact of Covid-19 pandemic on crime rate in Canada, focusing on the release of major covid-related policies;
Performed data cleaning, grouping and EDA using MySQL, labeled and related multiple dataset to accommodate further modelling;
Developed an interactive graphical user interface using Python Tkinter and Matplotlib to visualize the data according to user selected filters.

Python, Tkinter, Matplotlib, NumPy, Pandas, MySQL, Git

Used Twitter Developer API to collect a set of 30,000 tweets, cleaned the text by tokenizing and stemming, removed 120+ common stop words;
Vectorized the lists of words and fitted Naive Bayes models on the dataset, the best model achieved a test accuracy of 75.28%;
Debugged the model by adding Laplace Smoothing, which prevented the probabilities of zero;
Improved the model by adding TF-IDF and use lemmatizer instead of stemmer, which increase test accuracy by 5% and reduced overfitting.

Twitter Developer API, Python, NumPy, SciPy, Pandas, Matplotlib, nltk, sklearn, Git

Aimed to explore the commonly discussed topics that include the #cdnpoli hashtag.
Using data obtained from the Twitter Developer API, employed the Latent Dirichlet Allocation (LDA) to train model.
Identified 10 topics(word bags).
Analyzed the potential bias of the data, and limitations in LDA model such as challenges to interpret the key words of categorization, as the loss of imformation captured by emoji, images and outside URL links.

Twitter Developer API, Python, NumPy, SciPy, Pandas, Matplotlib, nltk, sklearn, Git, WordCloud, Gensim

Ongoing Projects...