Stock price prediction using probabilistic machine learning models

Ocepek, David (author), Demšar, Jure (mentor)

Our thesis had three main goals. The first one was to create a neural network-based stock portfolio trading agent for predicting future stock price distributions. This allows traders not only to estimate future trends for a given stock but also to predict risks associated with the stock. The second goal was to improve this agent's profitability by training it with custom loss functions, specifically designed for the nature of stock forecasting problems. The third goal was to create a realistic brokerage environment for running trading simulations in order to evaluate the success of our models. We developed several neural network-based models and compared them with more traditional time series forecasting approaches. For estimating uncertainty we used Markov chain dropout or the quantile loss. To further improve our results, agents facilitated two kinds of portfolio optimization techniques, the expected return maximization, which did not use uncertainty in its decision-making, and the chance-constrained optimization, which did. Based on our experiments, we conclude that trading commissions strongly affect agent profitability. The most profitable models were the temporal fusion transformer and a simple neural network, resulting in gains of 65% for US residents and 19% for non-US residents. The custom loss function did not yield the desired results and performed worse than mean square error. In terms of optimizers, chance-constrained optimization increases the profitability of agents but is riskier as it is unable to guarantee a certain level of risk to the user.

• Language: English
• Keywords: stock price forecasting, chance-constrained optimization, risk-estimation, stock trading, neural networks
• Work type: Master's thesis/paper
• Typology: 2.09 - Master's Thesis
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2023
• PID: 20.500.12556/RUL-152756
• COBISS.SI-ID: 177588483
• Publication date in RUL: 06.12.2023

OCEPEK, David, 2023, Stock price prediction using probabilistic machine learning models [online]. Master’s thesis. [Accessed 18 April 2025]. Retrieved from: https://repozitorij.uni-lj.si/IzpisGradiva.php?lang=eng&id=152756

Secondary language

• Language: Slovenian
• Title: Napovedovanje cen delnic z uporabo modelov verjetnostnega strojnega učenja

Abstract

Naša naloga je imela tri glavne cilje. Prvi cilj je bil ustvariti agenta za portfeljsko trgovanje z delnicami, ki bi uporabljal nevronske mreže za napovedovanje porazdelitev cen delnic. To bi omogočilo trgovalcem, ne le da ocenijo prihodnje trende v ceni določene delnice, ampak tudi da napovejo, kolikšno je tveganje, povezano s to delnico. Drugi cilj je bil izboljšati dobičkonosnost tega agenta z uporabo izgub, posebej zasnovanih za napovedovanje cen delnic. Tretji cilj je bil ustvariti realistično borzno okolje za simuliranje trgovanja, s pomočjo katerega smo ocenili uspešnost naših modelov. Razvili smo več nevronskih mrež in jih primerjali z bolj tradicionalnimi pristopi za napovedovanje časovnih nizov. Za oceno negotovosti smo uporabili Monte-Carlo dropout ali kvantilno funkcijo izgube. Za nadaljnje izboljšanje rezultatov so agenti uporabljali dve tehnike za optimizacijo portfelja: optimizacija pričakovanega donosa, ki ni upoštevala negotovosti pri sprejemanju odločitev, in optimizacija z verjetnostnimi pogoji, ki jo je upoštevala. Na podlagi naših eksperimentov sklepamo, da provizije za trgovanje močno vplivajo na dobičkonosnost agenta. Najbolj dobičkonosni modeli so bili časovni fuzijski transformator in preprosta nevronska mreža, z uporabo katerih je agent dosegel dobiček v višini 65% za rezidente ZDA in 19% za rezidente drugih držav. Prilagojena funkcija izgube je bila slabša od srednje kvadratne napake. Kar zadeva optimizatorje, je optimizacija z verjetnostnimi pogoji povečala dobičkonosnost agentov, vendar je bolj tvegana, saj ni mogla zagotoviti določene ravni tveganja za uporabnika.

• Keywords: napovedovanje cen delnic, optimizacija z verjetnostnimi omejitvami, ocena tveganja, trgovanje z delnicami, nevronske mreže