PREDICTING THE FUTURE DEVELOPMENT OF PANCREATIC CANCER

Pancreatic cancer is the fourth leading cause of death from cancer around the world. Factors known to increase the risk of developing pancreatic cancer include obesity, diabetes, physical activity, alcohol use, tobacco use, age, race, gender, and genetic exposure. The overall survival rate of pancreatic cancer is 5%, as pancreatic cancer is relatively asymptomatic until later stages, making individual knowledge of risk crucial. Neural networks and logistic regression are commonly used techniques for predicting outcomes. Artificial neural networks are a method of machine learning based on human learning and the structure of neurons. Logistic regression implements use of a logistic function to model dependent variables. The intended purpose of this experiment was to compare the effectiveness of using an artificial neural network in comparison to using logistic regression in order to predict the diagnosis of pancreatic cancer given certain factors about individuals. It was thought that due to the more individualized nature of the neural network, it will be more effective. Both methods were run with the same data set which contains approximately 10,000 individuals surveyed with multiple variables that may be linked with development of pancreatic cancer and accuracy was tested. Through this process, it was observed that neither were significantly different in effectiveness to predict the development of pancreatic cancer.