Fine Arts : 215
Date & Time
November 8, 2023, 11:00 am – 12:00 pm
|Session Chair:||Emi Galambos|
Speaker 1: Nathan Tamiru
- Epidemic Modeling Using the SIR Model and Graph Networks
- We present a method for simulating the spread of disease in an epidemic using a graph network framework based on the Susceptible-Infected-Recovered (SIR) model. In particular, we model disease propagation across a spatial region composed of distinct geographical units. We analyze the percentage of susceptible, infected, and recovered populations within each region over specified time intervals. Key inputs for our simulations include disease parameters and an adjacency matrix, which represents the mobility of individuals from one node to another. In our simulations, we mimic agent mobility within the network by transforming the adjacency matrix into a matrix which represents the relationships and connections between nodes in the network. We call this matrix the graph-Laplacian matrix. Using this Laplacian matrix and the specified disease parameters, we generate SIR graphs for each spatial region. Additionally, we perform a comparative analysis of errors arising from explicit and implicit solvers, through a series of numerical experiments. This analysis sheds light on the dynamics of disease propagation within a networked environment and offers insights into this model's performance and accuracy.
Speaker 2: Fred Azizi
- A comparison of machine learning models for time series forecasting
- We compare prominent machine learning models for time series forecasting, using Bitcoin price data as a case study. Our analysis encompasses a range of techniques, including Multilayer Perceptron (MLP), K-Nearest Neighbor Regression, Random Forest, and Support Vector Regression. We compare the performance of these machine learning models against the classical autoregressive integrated moving average (ARIMA) model. This comparative analysis takes into account predictive accuracy on a test dataset, utilizing the Symmetric Mean Absolute Percentage Error (SMAPE), as a key performance metric. We provide insights into the suitability of various machine learning models for time series forecasting, shedding light on their relative strengths and weaknesses in the context of cryptocurrency price prediction.