Predicting the price of petrochemical industry stocks using price action patterns and trading board insights

Abstract: Reducing transaction costs and accurately predicting stock prices are critical concerns for traders and market participants, particularly within Iran’s capital market. To address these challenges, a variety of models and methods have been employed, among which market effect models—especially those capturing immediate price effects—are widely recognized. The price effect refers to the change in a stock’s price following a transaction, and this study leverages such models to forecast price movements in the petrochemical sector. To achieve this, stocks within the petrochemical industry were first ranked by market value, and four representative stocks were selected using specific filtering criteria. To enhance model accuracy, transactions were categorized into buyer-initiated and seller-initiated groups. Price modeling was conducted using microstructural parameters in conjunction with temporal variables, applying the Heterogeneous Autoregressive (HAR) method. Findings indicate that the integration of time-based variables—such as days of the week—with microstructural data does not significantly improve model performance. In fact, temporal additions did not contribute to error reduction. Nonetheless, the HAR model outperforms alternatives cited in prior literature, demonstrating a meaningful decrease in prediction error on out-of-sample data. Keywords: Price Effect, Market Microstructure, Petrochemical Industry, Heterogeneous Autoregressive Model