Dr. Jesse Louis Barlow, Thesis Supervisor Dr. John Joseph Hannan, Thesis Honors Advisor
Keywords:
trading statistical modeling data mining machine learning
Abstract:
The thesis studies the effect of weekly search volume data from Google Trends on volatility measures of a portfolio of hand-picked stocks. Twelve stocks were selected from three sectors and a Granger causality analysis was performed to determine whether the search volume time series was useful in forecasting the volatility time series for a given stock. The re- sults from the Granger causality analysis showed that some, but not all, stocks could use their search volume data from Google Trends to signifi- cantly forecast their volatility. For those stocks whose search volume data proved fruitful in forecasting their volatility, a search volume model con- sisting of lags of search volume data as predictors was compared to a null model consisting of the average of the volatility as a forecast. Using the mean absolute percentage error as a metric, the results support the view that the search volume model does have some forecast ability in produc- ing volatility estimates.
