REVERSING CELLULAR AUTOMATA FOR BIOLOGICAL EVOLUTION
â–ª Professor Supervisor: Danny Barash
Researched the inverse problem in cellular automata to model biological evolution under the mentorship of Dr. Danny Barash and Dr. Ram M. Programmed evolutionary algorithms to identify rule mutation chains and analyze their fitness and lifetime dynamics, leveraging computational frameworks from Stephen Wolfram’s work to uncover emergent behavior in discrete dynamical systems.
PREDICTING FOOD INSECURITY THROUGH TEMPORAL AND SPATIAL ANALYSIS
Supervisor: Clayton Greenburg
Journal: GeoJournal
This research is submitted to be published in GeoJournal. Food insecurity is characterized by an insufficient food supply. In the United States, it impacts 47 million people annually, including 14 million children. Accurate predictive models are crucial for detecting challenges, distributing resources, and mitigating human suffering. This study presents three artificial intelligence models—autoregressive moving average (ARIMA), long short-term memory (LSTM), and convolutional neural network (CNN)—for predicting food insecurity at the county level. All models were trained with multiple time-series socioeconomic data from public resources obtained from 2018 to 2023 after cleaning, organization, and optimization. Food security served as a dependent variable, and 10 socioeconomic factors across 3,142 counties served as independent variables. CNN used heatmap-based geographical data, and the result indicates a significant correlation between unemployment and high number of single-parent households with food security. CNN achieved the lowest mean absolute error with the highest accuracy, followed by LSTM, and then ARIMA. Future research should consider more social and environmental factors to enhance geographic coverage and accuracy.
COMBATING CLOTHING INSECURITY THROUGH ARTIFICIAL INTELLIGENCE
This research aims to address clothing insecurity in the USA by leveraging artificial intelligence (AI) to predict the intensity of clothing insecurity and project clothing needs. By analyzing socio-economic data, weather patterns, natural disaster occurrences, and poverty metrics, the AI models will identify regions with clothing assistance requirements. Leveraging my technical expertise in Python, AWS, React Native, and AI, I developed a mobile app. This app integrates predictive modeling with user-friendly assessment and donation tools to facilitate efficient and impactful clothing distribution. With making Auto-Regressive-Integrated-Moving-Average(ARIMA) models for each socio-economic factor and time series models that uses data on socio-economic factors from previous years to project data for following years into the future, this app is able to provide a clothing insecurity index for the next year and informs users of the app to donate to the areas vulnerable to this issue. Utilizing a RandomForestClassifier, a classification system that fits a group of decision trees across sub-samples of a dataset and uses averaging to improve the predictive accuracy and control over-fitting, the number of clothing needed by age and gender is generated and informs users how much clothing is needed in the area. Results of the models indicate that the RMSE value of 0.0472. The app also provides options for recording donations for users of the app. With the lack of attention in donations for clothing insecurity, the mobile app, CARECloset_, diverts attention to a problem and inspires people to donate clothing. This app can be used as a solution to combating clothing insecurity using modelling techniques in order to provide a clothing insecurity index and amount of clothing needed in a specific county.