nep-for New Economics Papers
on Forecasting
Issue of 2022‒10‒24
five papers chosen by
Rob J Hyndman
Monash University

  1. Forecasting Regional Industrial Production with High-Frequency Electricity Consumption Data By Robert Lehmann; Sascha Möhrle
  2. Monotonic Neural Additive Models: Pursuing Regulated Machine Learning Models for Credit Scoring By Dangxing Chen; Weicheng Ye
  3. Predict stock prices with ARIMA and LSTM By Ruochen Xiao; Yingying Feng; Lei Yan; Yihan Ma
  4. Interpretable Selective Learning in Credit Risk By Dangxing Chen; Weicheng Ye; Jiahui Ye
  5. Physics-Informed Convolutional Transformer for Predicting Volatility Surface By Soohan Kim; Seok-Bae Yun; Hyeong-Ohk Bae; Muhyun Lee; Youngjoon Hong

  1. By: Robert Lehmann; Sascha Möhrle
    Abstract: In this paper, we study the predictive power of electricity consumption data for regional economic activity. Using unique weekly and monthly electricity consumption data for the second-largest German state, the Free State of Bavaria, we conduct a pseudo out-of-sample forecasting experiment for the monthly growth rate of Bavarian industrial production. We find that electricity consumption is the best performing indicator in the nowcasting setup and has higher accuracy than other conventional indicators in a monthly forecasting experiment. Exploiting the high-frequency nature of the data, we find that the weekly electricity consumption indicator also provides good predictions about industrial activity in the current month even with only one week of information. Overall, our results indicate that regional electricity consumption offers a promising avenue to measure and forecast regional economic activity.
    Keywords: electricity consumption, real-time indicators, forecasting, nowcasting
    JEL: E17 E27 R11
    Date: 2022
  2. By: Dangxing Chen; Weicheng Ye
    Abstract: The forecasting of credit default risk has been an active research field for several decades. Historically, logistic regression has been used as a major tool due to its compliance with regulatory requirements: transparency, explainability, and fairness. In recent years, researchers have increasingly used complex and advanced machine learning methods to improve prediction accuracy. Even though a machine learning method could potentially improve the model accuracy, it complicates simple logistic regression, deteriorates explainability, and often violates fairness. In the absence of compliance with regulatory requirements, even highly accurate machine learning methods are unlikely to be accepted by companies for credit scoring. In this paper, we introduce a novel class of monotonic neural additive models, which meet regulatory requirements by simplifying neural network architecture and enforcing monotonicity. By utilizing the special architectural features of the neural additive model, the monotonic neural additive model penalizes monotonicity violations effectively. Consequently, the computational cost of training a monotonic neural additive model is similar to that of training a neural additive model, as a free lunch. We demonstrate through empirical results that our new model is as accurate as black-box fully-connected neural networks, providing a highly accurate and regulated machine learning method.
    Date: 2022–09
  3. By: Ruochen Xiao; Yingying Feng; Lei Yan; Yihan Ma
    Abstract: MAE, MSE and RMSE performance indicators are used to analyze the performance of different stocks predicted by LSTM and ARIMA models in this paper. 50 listed company stocks from are selected as the research object in the experiments. The dataset used in this work consists of the highest price on transaction days, corresponding to the period from 01 January 2010 to 31 December 2018. For LSTM model, the data from 01 January 2010 to 31 December 2015 are selected as the training set, the data from 01 January 2016 to 31 December 2017 as the validation set and the data from 01 January 2018 to 31 December 2018 as the test set. In term of ARIMA model, the data from 01 January 2016 to 31 December 2017 are selected as the training set, and the data from 01 January 2018 to 31 December 2018 as the test set. For both models, 60 days of data are used to predict the next day. After analysis, it is suggested that both ARIMA and LSTM models can predict stock prices, and the prediction results are generally consistent with the actual results;and LSTM has better performance in predicting stock prices(especially in expressing stock price changes), while the application of ARIMA is more convenient.
    Date: 2022–08
  4. By: Dangxing Chen; Weicheng Ye; Jiahui Ye
    Abstract: The forecasting of the credit default risk has been an important research field for several decades. Traditionally, logistic regression has been widely recognized as a solution due to its accuracy and interpretability. As a recent trend, researchers tend to use more complex and advanced machine learning methods to improve the accuracy of the prediction. Although certain non-linear machine learning methods have better predictive power, they are often considered to lack interpretability by financial regulators. Thus, they have not been widely applied in credit risk assessment. We introduce a neural network with the selective option to increase interpretability by distinguishing whether the datasets can be explained by the linear models or not. We find that, for most of the datasets, logistic regression will be sufficient, with reasonable accuracy; meanwhile, for some specific data portions, a shallow neural network model leads to much better accuracy without significantly sacrificing the interpretability.
    Date: 2022–09
  5. By: Soohan Kim; Seok-Bae Yun; Hyeong-Ohk Bae; Muhyun Lee; Youngjoon Hong
    Abstract: Predicting volatility is important for asset predicting, option pricing and hedging strategies because it cannot be directly observed in the financial market. The Black-Scholes option pricing model is one of the most widely used models by market participants. Notwithstanding, the Black-Scholes model is based on heavily criticized theoretical premises, one of which is the constant volatility assumption. The dynamics of the volatility surface is difficult to estimate. In this paper, we establish a novel architecture based on physics-informed neural networks and convolutional transformers. The performance of the new architecture is directly compared to other well-known deep-learning architectures, such as standard physics-informed neural networks, convolutional long-short term memory (ConvLSTM), and self-attention ConvLSTM. Numerical evidence indicates that the proposed physics-informed convolutional transformer network achieves a superior performance than other methods.
    Date: 2022–09

This nep-for issue is ©2022 by Rob J Hyndman. It is provided as is without any express or implied warranty. It may be freely redistributed in whole or in part for any purpose. If distributed in part, please include this notice.
General information on the NEP project can be found at For comments please write to the director of NEP, Marco Novarese at <>. Put “NEP” in the subject, otherwise your mail may be rejected.
NEP’s infrastructure is sponsored by the School of Economics and Finance of Massey University in New Zealand.