Multi-Source Fusion-Based Model with Customized Loss Function for Stock Selection
Selecting outstanding tech stocks for investment is challenging for any market investor. This challenge comes from the complexity caused by the dynamic and volatile nature of public policies and macroeconomic factors. Moreover, this task appears complex as an investor needs to decide from a diverse set of stocks, where performance ranking can conflict with another for different criteria, e.g., the contradiction between pure returns and risk performance. The paper proposes a novel model for stock evaluation and selection problems based on multi-source data fusion and decision-level fusion by multi-criteria decision-making (MCDM) techniques. The narrow criteria performance of a single source usually restricts a candidate's overall evaluation, especially on stock selection. Therefore, the proposed multi-source acquisition is promising to enhance reliability with reasonable data fusion, Furthermore, proper raw data processing and information integration by data fusion techniques are essential to obtain more accurate and concise data for decision-making support. Most important, the proposed multi-MCDM decision-fusion procedure can enhance the strength of a model to overcome the weakness of a single-performed model. Besides, we propose a customized loss function that improves the accuracy of time-series data prediction and increases the data's credibility. Meanwhile, the experiment shows that the proposed method reduces the prediction log error by 28\% in a best case on the experimental real-situation data and decreases the time-series data matching cost by 20\% between the prediction and observation. Finally, we conducted the real-world stock data selection by the proposed fusion-based model. The proposed model locates five potential preferred stocks, and the results are practical and effective, which are further justified through a detailed ablation study. Meanwhile, the time-series forecasting data got a lower error prediction than the baseline loss
Year of publication: |
2023
|
---|---|
Authors: | Snasel, Vaclav ; Domingo Velasquez Silva, Juan ; Pant, Millie ; Georgiou, Dimitrios ; kong, lingping |
Publisher: |
[S.l.] : SSRN |
Saved in:
freely available
Saved in favorites
Similar items by person
-
Alwahaishi, Saleh, (2014)
-
Alwahaishi, Saleh, (2014)
-
Limiting Fiscal Procyclicality : Evidence from Resource-Rich Countries
Coutinho, Leonor, (2020)
- More ...