Working Papers

The WNE Working Papers series has been published by the Faculty of Economic Sciences at the University of Warsaw since 2008.

The WNE Working Papers series provides a fast, open channel for disseminating research conducted at the Faculty of Economic Sciences, University of Warsaw. The papers hosted here are working versions (pre-prints) and may evolve as authors refine their analyses, incorporate feedback, or progress through formal peer-review. What you read is the current version released by the authors, timestamped and assigned DOI/ISSN identifiers to ensure precise citation and version tracking. Copyright remains with the authors, who may, at any time, upload a revised file or add a note directing readers to a later, peer-reviewed publication.

The Working Papers series accepts articles by research employees of the Faculty and publications from conferences organised at the Faculty of Economic Sciences at the University of Warsaw. Articles should be original research papers which have not been previously published, on the subject of economics.

Please send your paper by e-mail: YTzG*d0V\Ziq'_/tCN31j?]wU$!]#[QDm/nSrG=KOdvv%c!W&&s{Bm^ql 

Please send 2 files:

(1) the main text without the title of the article and the authors (DOC/DOCX file) and

(2) the title page including: the title of the paper, the authors and their affiliation (DOC/DOCX file).

Please read the detailed editing requirements before submitting your text. 

Number of Results 6

WP(18/2026)512. Painting Price: A Machine Learning Approach to Art Valuation. Proof of Concept and Market Structure Diagnosis

Authors: Okhrimenko Kostiantyn
This paper investigates the feasibility of predicting art prices using machine learning methods applied to a dataset of 20,905 paintings and drawings scraped from the Artsper online marketplace. We test tree-based ensemble models (Decision Tree, Rand…

WP512. Painting Price: A Machine Learning Approach to Art Valuation. Proof of Concept and Market Structure Diagnosis

This paper investigates the feasibility of predicting art prices using machine learning methods applied to a dataset of 20,905 paintings and drawings scraped from the Artsper online marketplace. We test tree-based ensemble models (Decision Tree, Random Forest, XGboost) and deep learning architectures (MLP, CNN, Fusion) on both tabular metadata and hand-crafted image features. Results consistently show poor predictive performance across all model types and feature sets. We argue that this outcome is not a methodological failure but a substantive finding: it constitutes a diagnosis of the market structure of contemporary art. Drawing on hedonic pricing theory (Rosen 1974), the sociology of cultural fields (Bourdieu 1993), and the economics of valuation (Velthuis 2005; Beckert and Rössel 2013), we propose a three-layer model of art price determinants: physical attributes (observable and partially captured by models), visual-aesthetic features (observable but poorly quantifiable), and narrative-reputational capital (largely unobservable in cross-sectional platform data).


DOI: https://doi.org/10.33138/2957-0506.2026.18.512
Okhrimenko Kostiantyn
art market machine learning hedonic pricing art valuation cultural economics XGboost CNN

WP(17/2023)424. Optimal Markowitz Portfolio Using Returns Forecasted with Time Series and Machine Learning Models

Authors: Ślusarczyk Damian, Ślepaczuk Robert
We aim to answer the question of whether using forecasted stock returns based on machine learning and time series models in a mean-variance portfolio framework yields better results than relying on historical returns. Nevertheless, the problem of the…

WP424. Optimal Markowitz Portfolio Using Returns Forecasted with Time Series and Machine Learning Models

We aim to answer the question of whether using forecasted stock returns based on machine learning and time series models in a mean-variance portfolio framework yields better results than relying on historical returns. Nevertheless, the problem of the efficient stock selection has been tested for more than 50 years, the issue of adequate construction of mean-variance portfolio framework and incorporating forecasts of returns in it has not been solved yet. Stock returns portfolios were created using ’raw’ historical returns and forecasted return based on ARIMA-GARCH and the XGBoost models. Two optimization problems were concerned: global maximum information ratio and global mini-mum variance. Then strategies were compared with two benchmarks – an equally weighted portfolio and buy and hold on the DJIA index. Strategies were tested on Dow Jones Industrial Average stocks in the period from 2007-01-01 to 2022-12-31 and daily data was used. The main portfolio performance metrics were information ratio* and information ratio**. The results showed that using forecasted returns we can enhance our portfolio selection based on Markowitz framework, but it is not a universal solution, and we have to control all the parameters and hyperparameters of selected models.


Ślusarczyk Damian, Ślepaczuk Robert
Algorithmic Investment Strategies Markowitz framework portfolio optimization forecasting ARIMA GARCH XGBoost minimum variance

WP(15/2023)422. Ensembled LSTM with Walk Forward Optimization in Algorithmic Trading

Authors: Chojnacki Karol, Ślepaczuk Robert
This study compares well-known tools of technical analysis (Moving Average Crossover MAC) with Machine Learning based strategies (LSTM and XGBoost) and Ensembled Machine Learning Strategies (LSTM ensembled with XGBoost and MAC). All models were …

WP422. Ensembled LSTM with Walk Forward Optimization in Algorithmic Trading

This study compares well-known tools of technical analysis (Moving Average Crossover MAC) with Machine Learning based strategies (LSTM and XGBoost) and Ensembled Machine Learning Strategies (LSTM ensembled with XGBoost and MAC). All models were compared to Buy and Hold benchmark and evaluated using Performance Metrics, that is Annualized Return Compounded, Maximum Drawdown, Maximum Loss Duration, and three types of Information Ratio. This research uses daily S&P 500 index data ranging from 2000 to 2023. Every strategy was optimized with novel walk forward approach consisting of numerous in sample and out of sample periods. MAC and best performing ML methods were subjected to sensitivity analysis. The results show that LSTM ensembled with XGBoost and MAC yields the most promising results in terms of risk-adjusted returns which suggest further research focused on ensembling of individual ML strategies. Finally, we show that classical methods of technical analysis (that is, MAC) are much less robust and indifferent to change in hyperparameters than machine learning based algorithms, especially LSTM.


Chojnacki Karol, Ślepaczuk Robert
Algorithmic Investment Strategies Machine Learning Recurrent Neural Networks Long Short-Term Memory XGBoost Walk Forward Optimization Trading algorithms Technical Analysis Indicators

WP(13/2023)420. From Alchemy to Analytics: Unleashing the Potential of Technical Analysis in Predicting Noble Metal Price Movement

Authors: Chlebus Marcin, Nowak Artur
Algorithmic trading has been a central theme in numerous research papers, combining knowledge from the fields of Finance and Mathematics. This thesis aimed to apply basic Technical Analysis indicators for predicting price movement of three noble meta…

WP420. From Alchemy to Analytics: Unleashing the Potential of Technical Analysis in Predicting Noble Metal Price Movement

Algorithmic trading has been a central theme in numerous research papers, combining knowledge from the fields of Finance and Mathematics. This thesis aimed to apply basic Technical Analysis indicators for predicting price movement of three noble metals: Gold, Silver, and Platinum in a form of multi-class classification. That task was performed using four algorithms: Logistic Regression, k-Nearest Neighbors, Random Forest and XGBoost. The study incorporated feature filtering methods such as Kendall-tau filtering and PCA, as well as five different data frequencies: 1, 5, 10, 15 and 20 trading days. From a total of 40 potential models for each metal, the best one was selected and evaluated using data from period 2018-2022. The result revealed that models utilizing only Technical Analysis indicators were able to predict price movements to a significant extent, leading to investment strategies that outperformed the market in two out of three cases.


Chlebus Marcin, Nowak Artur
precious metals algotrading machine learning multiclass classification logistic regression nearest neighbors random forest xgboost

WP(22/2021)370. Predicting football outcomes from Spanish league using machine learning models

Authors: Lewandowski Michał, Chlebus Marcin
High-quality football predictive models can be very useful and profitable. Therefore, in this research, we undertook to construct machine learning models to predict football outcomes in games from Spanish LaLiga and then we compared them with histori…

WP370. Predicting football outcomes from Spanish league using machine learning models

High-quality football predictive models can be very useful and profitable. Therefore, in this research, we undertook to construct machine learning models to predict football outcomes in games from Spanish LaLiga and then we compared them with historical forecasts extracted from bookmakers, which knowledge is commonly considered to be deep and high-quality. The aim of the paper was to design models with the highest possible predictive performances, get results close to bookmakers or even building better estimators. The work included detailed feature engineering based on previous achievements of this domain and own proposals. A built and selected set of variables was used with four machine learning methods, namely Random Forest, AdaBoost, XGBoost and CatBoost. The algorithms were compared based on: Area Under the Curve (AUC) and Ranked Probability Score (RPS). RPS was used as a benchmark in the comparison of estimated probabilities from trained models and forecasts from bookmakers' odds. For a deeper understanding and explanation of the demonstrated methods, which are considered as black-box approaches, Permutation Feature Importance (PFI) was used to evaluate the impacts of individual variables. Features extracted from bookmakers odds’ occurred the most important in terms of PFI. Furthermore, XGBoost achieved the best results on the validation set (RPS equals 0.1989), which obtained similar predictive power to bookmakers' odds (their RPS between 0.1977 and 0.1984). Results of the trained estimators were promising and this article showed that competition with bookmakers is possible using demonstrated techniques.


Lewandowski Michał, Chlebus Marcin
predicting football outcomes machine learning betting adaboost random forest xgboost catboost ranked probability score auc permutation feature importance

WP(15/2020)321. Comparison of tree-based models performance in prediction of marketing campaign results using Explainable Artificial Intelligence tools

Authors: Chlebus Marcin, Osika Zuzanna
The research uses tree-based models to predict the success of telemarketing campaign of Portuguese bank. The Portuguese bank dataset was used in the past in different researches with different models to predict the success of campaign. We propose to …

WP321. Comparison of tree-based models performance in prediction of marketing campaign results using Explainable Artificial Intelligence tools

The research uses tree-based models to predict the success of telemarketing campaign of Portuguese bank. The Portuguese bank dataset was used in the past in different researches with different models to predict the success of campaign. We propose to use boosting algorithms, which have not been used before to predict the response for the campaign and to use Explainable AI (XAI) methods to evaluate model's performance. The paper tries to examine whether 1) complex boosting algorithms perform better and 2) XAI tools are better indicators of models' performance than commonly used discriminatory power's measures like AUC. Portuguese bank telemarketing dataset was used with five machine learning algorithms, namely Random Forest (RF), AdaBoost, GBM, XGBoost and CatBoost, which were then later compared based on their AUC and XAI tools analysis – Permutated Variable Importance and Partial Dependency Profile. Two best performing models based on their AUC were XGBoost and CatBoost, with XGBoost having slightly higher AUC. Then, these models were examined using PDP and VI, which resulted in discovery of XGBoost potenitial overfitting and choosing CatBoost over XGBoost. The results show that new boosting models perform better than older models and that XAI tools could be helpful with models' comparisons.


Chlebus Marcin, Osika Zuzanna
direct marketing telemarketing relationship marketing data mining machine learning random forest adaboost gbm catboost xgboost bank marketing XAI variable importance partial dependency profile
  • 1