Measuring financial soundness around the world: A machine learning approach

We use a fully data-driven approach and information provided by the IMF’s financial soundness indicators to measure the condition of a country’s financial system around the world. Given the nature of the measurement problem, we apply different versions of principal component analysis (PCA) to deal with the presence of strong cross-sectional and time dependence in the data due to unobserved common factors. Using this comprehensive sample and various statistical methods, we produce an alternative data-driven measure of financial soundness that provides policy makers and financial institutions with a monitoring and policy tool that is easy to implement and update. We validate our index by using alternative macroeconomic factors, confirming its predictive power. Our index captures important aspects of financial intermediation around the world.     

A random forest-based model for crypto asset forecasts in futures markets with out-of-sample prediction

In this study, a price prediction model for futures markets of crypto assets is presented. Random Forest was used to study three scenarios as a function of input variables: technical indicators, candlestick patterns and both simultaneously. In turn, the model parameters, the time intervals, and the most suitable investment horizons were studied. In addition to showing the results from the model, a one-year out-of-sample prediction was simulated. The entire year of 2020 was chosen because the three possible stock market scenarios occurred in this year: a sideways market, a bear market resulting from the global pandemic and an end-of-year bull market. Last, this out-of-sample simulation was analyzed as a real operation, that is, by retraining the model after each new collection of data, so that the model had the maximum information at all times. In conclusion, using candlestick patterns instead of technical indicators, improves the efficiency of the results.     

Multilevel IRT models for the analysis of satisfaction for distance learning during the Covid-19 pandemic

The Covid-19 pandemic played a relevant role in the diffusion of distance learning alternatives to “traditional” learning based on classroom activities, to allow university students to continue attending lessons during the most severe phases of the pandemic. In such a context, investigating the students' perspective on distance learning provides useful information to stakeholders to improve effective educational strategies, which could be useful also after the end of the emergency to favor the digital transformation in the higher educational setting.Here we focus on the satisfaction in distance learning for Italian university students. We rely on data comprising students enrolled in various Italian universities, which were inquired about several aspects related to learning distance.We explicitly take into account the hierarchical nature of data (i.e., students nested in universities) and the latent nature of the variable of interest (i.e., students' learning satisfaction) through a multilevel Item Response Theory model with students' and universities' covariates.As the main results of our study, we find out that distance learning satisfaction of students: (i) depends on the University where they study; (ii) is affected by some students' socio-demographic characteristics, among which psychological factors related to Covid-19; (iii) is affected by some observable university characteristics.     

Machine learning and the cross-section of emerging market stock returns

This paper compares various machine learning models to predict the cross-section of emerging market stock returns. We document that allowing for non-linearities and interactions leads to economically and statistically superior out-of-sample returns compared to traditional linear models. Although we find that both linear and machine learning models show higher predictability for stocks associated with higher limits to arbitrage, we also show that this effect is less pronounced for non-linear models. Furthermore, significant net returns can be achieved when accounting for transaction costs, short-selling constraints, and limiting our investment universe to big stocks only.     

Business model contributions to bank profit performance: A machine learning approach

This paper examines the relation between bank profit performance and business models, using a machine learning–based approach. The analysis contributes to the literature on this relation by considering the bank portfolio’s ability to yield profits as the identification criterion of strategic profiles and by including all the components of the business model simultaneously in the identification process. Our research strategy is applied to the European Union banking system from 1997 to 2021. The paper’s primary finding indicates that specialization seems to be a strategy that results in banks adopting business profiles with better profit performance, particularly if the banks specialize in the standard retail-oriented model.     

专利转化特征精准识别与预测——以人工智能芯片为例

探讨前沿科技领域专利转化特征并对其进行精准识别与预测，对于我国破解“卡脖子”技术难题及实现科技自立自强具有重要意义。选取人工智能芯片专利领域，采用机器学习算法测度最优转化预测方案，分析全球范围内主要国家或地区专利成功转化影响因素，从企业/高校、国内/国际等不同层面总结专利成功转化的主要特征。结果发现：随机森林算法预测效果较好，人工智能芯片领域专利转化概率服从对数曲线分布，影响高校/企业、国内/国外专利转化特征的因素有所不同。最后，提出高校/科研机构应注重高价值专利维持和团队合作、企业应提升专利技术质量和撰写质量等政策建议。     

Measuring informational efficiency of the European carbon market — A quantitative evaluation of higher order dependence

This paper introduces a new method for measuring nonlinear predictability in financial price changes: the so-called intermittency coefficient, a parameter of the multifractal random walk model by Bacry et al. (2001). As the intermittency coefficient can quantify the degree of nonlinear deviation from a random walk, we employ its estimates from financial data as a proxy for the loss of financial market efficiency. In addition, we propose a new statistical test of the random walk hypothesis. In an empirical application using data from the largest currently existing market for tradable pollution permits, the European Union Emissions Trading Scheme (EU ETS), we show that the degree of efficiency of this market remains largely unchanged over the period of observation 2008–2019. This suggests that the market has reached a mature state: informational efficiency in Phase III remains at a level comparable to Phase II. What is more, the EU ETS is found to be more efficient than the US stock market. This result, surprising as such, is largely attributable to the lower exposure to global economic shocks of the EU ETS.     

Forecasting solar stock prices using tree-based machine learning classification: How important are silver prices?

Solar energy is one of the fastest growing sources of electricity generation. Forecasting solar stock prices is important for investors and venture capitalists interested in the renewable energy sector. This paper uses tree-based machine learning methods to forecast the direction of solar stock prices. The feature set used in prediction includes a selection of well-known technical indicators, silver prices, silver price volatility, and oil price volatility. The solar stock price direction prediction accuracy of random forests, bagging, support vector machines, and extremely randomized trees is much higher than that of logit. For a forecast horizon of between 8 and 20 days, random forests, bagging, support vector machines, and extremely randomized trees achieve a prediction accuracy greater than 85%. Although not as prominent as technical indicators like MA200, WAD, and MA20, oil price volatility and silver price volatility are also important predictors. An investment portfolio trading strategy based on trading signals generated from the extremely randomized trees stock price direction prediction outperforms a simple buy and hold strategy. These results demonstrate the accuracy of using tree-based machine learning methods to forecast the direction of solar stock prices and adds to the broader literature on using machine learning techniques to forecast stock prices.     

Targeting predictors in random forest regression

Random forest (RF) regression is an extremely popular tool for analyzing high-dimensional data. Nonetheless, its benefits may be lessened in sparse settings due to weak predictors, and a pre-estimation dimension reduction (targeting) step is required. We show that proper targeting controls the probability of placing splits along strong predictors, thus providing an important complement to RF’s feature sampling. This is supported by simulations using finite representative samples. Moreover, we quantify the immediate gain from targeting in terms of the increased strength of individual trees. Macroeconomic and financial applications show that the bias–variance trade-off implied by targeting, due to increased correlation among trees in the forest, is balanced at a medium degree of targeting, selecting the best 5%–30% of commonly applied predictors. Improvements in the predictive accuracy of targeted RF relative to ordinary RF are considerable, up to 21%, occurring both in recessions and expansions, particularly at long horizons.