MODELLING MUTUAL INTERACTION OF FINANCE AND HUMAN FACTOR VIA VARIOUS SORTS OF INDICES

Date

2023-6-22

Author

Kalaycı, Betül

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

330
views

60
downloads

Cite This

This thesis represents the mutual effects between some financial processes and sentiment indices by using various models from machine learning approaches and nonparametric models to parametric volatility models. In the analyses, we compare the gain in accuracy and computational time. We also evaluate the forecasting performance of sentiment index, consumer confidence index, consumer price index, unemployment rate and currency rate. Hereby, initially, we use sole multivariate adaptive regression splines (MARS), neural network (NN) and random forest (RF) models. Then, we apply two-stage hybrid models, namely, MARS-NN, MARS-RF, RF-MARS, RF-NN, NN-MARS, and NN-RF. Finally, we implement volatility models for sentiment index and consumer confidence index, and investigate plausible relationships with the selected macroeconomic data to improve the performance of forecast. In the interpretation of the findings, as the underlying datasets are prone to exhibit significant structural breaks, we apply the Markov switching model, define the location of breaks and lastly, we perform distinct time series volatility models. The results indicate better accuracy under Markov switching generalized autoregressive conditional heteroskedastic model, shortly, MSGARCH, among alternatives.

Subject Keywords

Investor Sentiment, Consumer Confidence Index, Sentiment Index, Machine Learning, Volatility Model, Markov Switching Model

URI

https://hdl.handle.net/11511/104673

Collections

Graduate School of Applied Mathematics, Thesis