Forecasting Ability Research Articles

Monthly rainfall prediction for different climatic zones in south Africa for 2024 using a random forest model

This study predicted 2024 rainfall in different climatic zones in South Africa using a random forest model. Previous studies have shown that random forests performed better than other models for rainfall prediction. South Africa was divided into nine using the Koppen-Geiger climate classification system, and three cities were selected for each climatic zone. Atmospheric datasets from the South African Weather Service and the National Aeronautics and Space Agency were used for this study. The datasets were trained, tested, and validated to assess the model's accuracy. With good forecast ability observed, the random forest was then used for monthly rainfall for 2024. The result of this prediction was then compared with 2022 and 2023 rainfall. The result indicated months where much rain should be expected in various cities and cities that may likely experience droughts. This result is particularly important for agriculture, water resource management, and early drought/flooding warning systems.

Forecasting U.S. stock returns conditional on geopolitical risk and business cycles

Using standard predictors in the forecasting literature, we forecast the U.S. stock market returns conditional on geopolitical risk and business cycles over the 1927–2021 period. We find that out-of-sample forecasting performance is significantly better in times of high geopolitical risk versus low geopolitical risk. Consistent with previous research, we find further evidence of improved return predictability in recessions. However, we find little difference in forecasting performance in recessions versus expansions once the level of geopolitical risk is controlled for. We find similar results when using stock market cycles and periods of positive/negative industrial production growth in place of recessions/expansions. Our study contributes to the forecasting literature by documenting that geopolitical risk by itself and in combination with business cycle indicators impacts the forecasting ability of standard forecasting variables in the literature. We also contribute to the literature on the adaptive markets hypothesis with evidence of time-varying return predictability. We find inconclusive evidence as to whether our results are based on time-varying predictability or time-varying risk.

PREDIKSI PENJUALAN MELALUI METODE WEIGHTED MOVING AVERAGE (WMA) PADA UMKM “Nasi Bakar”

This study aims to find out the results of sales forecasting calculations using the Weight Moving Average method in Nurochmah Nasi Bakar MSMEs from April 2023 to July 2024. In addition, to determine the accuracy of forecast errors in sales forecasting calculations and the application of the Weighted Moving Average (WMA) method in predicting sales to Nurochmah MSMEs. The research method used is descriptive quantitative using historical sales data from January 2023 to July 2024, to predict sales using the WMA method at Nasi Bakar Nurochmah MSMEs from April 2023 to July 2024 using Production and Operations Management-Quality Control (POM-QM) software. The results of the calculation of sales forecasting using the WMA method on Nurochmah grilled rice MSMEs for April 2023 to July 2024 fluctuated as follows: December was 768.4 packs, a decrease of 4.547% from the actual 805 packs. January was 772.7 packs, a decrease of 0.553% from the actual 777 packs, February was 781.4 packs, an increase of 5.31% from the actual 742 packs. The accuracy of forecast error in calculating sales forecasting using the WMA method from April 2023 to July 2024 obtained an accuracy value of < 10% measured using the Mean Absolute Percent Error (MAPE). The average MAPE is 1.983%, it can be said that the forecasting ability of this method is very accurate. This shows that the WMA method can be applied to Nurochmah grilled rice MSMEs, so that it can help control sales volume, as well as as a basis for decision-making and planning sales targets for the next period.

Deep learning method for predicting weekly anatomical changes in patients with nasopharyngeal carcinoma during radiotherapy.

Patients may undergo anatomical changes during radiotherapy, leading to an underdosing of the target or overdosing of the organs at risk (OARs). This study developed a deep-learning method to predict the tumor response of patients with nasopharyngeal carcinoma (NPC) during treatment. This method can predict the anatomical changes of a patient. The participants included 230 patients with NPC. The data included planning computed tomography (pCT) and routine cone-beam CT (CBCT) images. The CBCT image quality was improved to the CT level using an advanced method. A long short-term memory network-generative adversarial network (LSTM-GAN) is proposed, which can harness the forecasting ability of LSTM and the generation ability of GAN. Four models were trained to predict the anatomical changes that occurred in weeks 3-6 and named LSTM-GAN-week 3 to LSTM-GAN-week 6. The pCT and CBCT were used as input, and the tumor target volumes (TVs) and OARs were delineated on the predicted and real images (ground truth). Finally, the models were evaluated using contours and dosimetry parameters. The proposed method predicted the anatomical changes, with a dice similarity coefficient above 0.94 and 0.90 for the TVs and surrounding OARs, respectively. The dosimetry parameters were close between the prediction and ground truth. The deviations in the prescription, minimum, and maximum doses of the tumor targets were below 0.5Gy. For serial organs (brain stem and spinal cord), the deviations in the maximum dose were below 0.6Gy. For parallel organs (bilateral parotid glands), the deviations in the mean dose were below 0.8Gy. The proposed method can predict the tumor response to radiotherapy in the future such that adaptation can be scheduled on time. This study provides a proactive mechanism for planning adaptation, which can enable personalized treatment and save clinical time by anticipating and preparing for treatment strategy adjustments.

Bayesian ex Post Evaluation of Recursive Multi-Step-Ahead Density Prediction

This research is focused on a formal Bayesian method of recursive multi-step-ahead density prediction and its ex post evaluation. Our approach remains within the framework of the standard (classical or orthodox) Bayesian paradigm based on the Bayes factor and on the use of the likelihood-based update. We propose a new decomposition of the predictive Bayes factor into the product of partial Bayes factors, for both a finite number of consecutive k-step-ahead forecasts (where k>1) and the recursive updates of the posterior odds ratios based on updated data sets. The first factor in the decomposition is related to the relative k-step-ahead forecasting ability of models, while the second one measures the updating effect. To illustrate the usefulness of the proposed measures, we apply the new decomposed predictive Bayes factors to compare the forecasting ability of models when the true data generating process (DGP) is known, using simulated data sets. Taking into account the effect of updating, the posterior odds ratios leads to the conclusion that the best model coincides with the true DGP. However, the highest k-step-ahead forecasting ability (considered alone) can be achieved by some other, less adequate models. Next, we investigate the predictive ability of different Vector Error Correction (VEC) models with conditional heteroscedasticity, combining three macroeconomic variables: unemployment, inflation and interest rates, separately for the US and Polish economies. The results show that the inference about the models’ predictive performance depends on the forecast horizon as well as on taking into account the updating effect.

A novel pathloss prediction and optimization approach using deep learning in millimeter wave communication systems

Millimeter wave communication systems design requires accurate pathloss prediction as well as optimization. Traditional pathloss models gives less accuracy therefore pathloss models using deep learning is alternative solution for traditional method. These methods overcome the time consuming pathloss prediction parameter prediction based on large dataset which is available for different frequencies. In this proposed work, for prediction of pathloss, Convolutional Neural Network (CNN) and for optimization of pathloss Particle Swarm Optimization (PSO) is proposed. Towards implementation of the same, outdoor scenario with two roads and one cross section is used and dataset is generated using DeepMIMO generic dataset. The simulation result shows that using optimization technique reduces pathloss and CNN give accurate prediction of 95.7 %. Furthermore, root mean square error value of 6.6 dB has been achieve which is less in comparison to alternative approaches designed for outdoor environments. The forthcoming development of multifrequency path loss models has sparked significant interest due to their forecasting ability to support higher frequency bands and address blockage scenarios within future wireless networks.

Improving the explainability of autoencoder factors for commodities through forecast-based Shapley values

Autoencoders are dimension reduction models in the field of machine learning which can be thought of as a neural network counterpart of principal components analysis (PCA). Due to their flexibility and good performance, autoencoders have been recently used for estimating nonlinear factor models in finance. The main weakness of autoencoders is that the results are less explainable than those obtained with the PCA. In this paper, we propose the adoption of the Shapley value to improve the explainability of autoencoders in the context of nonlinear factor models. In particular, we measure the relevance of nonlinear latent factors using a forecast-based Shapley value approach that measures each latent factor’s contributions in determining the out-of-sample accuracy in factor-augmented models. Considering the interesting empirical instance of the commodity market, we identify the most relevant latent factors for each commodity based on their out-of-sample forecasting ability.

Performance assessment of indicators of a multi-hazards early warning system in an urban mountain region

The urban basins in the mountainous Andean regions are exposed to rapid and poorly planned urbanization processes, causing drastic changes in the landscape, climate, and hydrology. These factors have led to an intensification in the frequency of floods, landslides, falling trees, and droughts, among many other disasters. Therefore, the city of Manizales has implemented an Early Warning System (EWS) to address emergencies and prevent deaths and damage from some of these disasters. However, these hazards can occur simultaneously, further complicating emergency response efforts and requiring the EWS to address multiple hazards; therefore, it has become the Multi-Hazards Early Warning System (MHEWS), which requires indicators and thresholds to generate alerts within sufficient time. The MHEWS implemented in Manizales addresses initially the hazards of floods and landslides, using indicators and thresholds related to recent rainfall and river levels. This article evaluates these proposed indicators by assessing their forecasting capability, effectiveness, and efficiency in increasing warning times, the alarm duration compared to the event duration, and their forecasting ability based on the location of rainfall triggering floods, landslides and other hazards identify. The analyzed indicators focus on hazards and do not consider exposure or vulnerability in communities. The results indicate that a single hazard indicator is not enough, and including multiple indicators based on complementary sources of information improves the performance of MHEWS in addressing emergencies caused by rain.

Daily Runoff Prediction Based on FA-LSTM Model

Accurate and reliable short-term runoff prediction plays a pivotal role in water resource management, agriculture, and flood control, enabling decision-makers to implement timely and effective measures to enhance water use efficiency and minimize losses. To further enhance the accuracy of runoff prediction, this study proposes a FA-LSTM model that integrates the Firefly algorithm (FA) with the long short-term memory neural network (LSTM). The research focuses on historical daily runoff data from the Dahuangjiangkou and Wuzhou Hydrology Stations in the Xijiang River Basin. The FA-LSTM model is compared with RNN, LSTM, GRU, SVM, and RF models. The FA-LSTM model was used to carry out the generalization experiment in Qianjiang, Wuxuan, and Guigang hydrology stations. Additionally, the study analyzes the performance of the FA-LSTM model across different forecasting horizons (1–5 days). Four quantitative evaluation metrics—mean absolute error (MAE), root mean square error (RMSE), coefficient of determination (R2), and Kling–Gupta efficiency coefficient (KGE)—are utilized in the evaluation process. The results indicate that: (1) Compared to RNN, LSTM, GRU, SVM, and RF models, the FA-LSTM model exhibits the best prediction performance, with daily runoff prediction determination coefficients (R2) reaching as high as 0.966 and 0.971 at the Dahuangjiangkou and Wuzhou Stations, respectively, and the KGE is as high as 0.965 and 0.960, respectively. (2) FA-LSTM model was used to conduct generalization tests at Qianjiang, Wuxuan and Guigang hydrology stations, and its R2 and KGE are 0.96 or above, indicating that the model has good adaptability in different hydrology stations and strong robustness. (3) As the prediction period extends, the R2 and KGE of the FA-LSTM model show a decreasing trend, but the whole model still showed feasible forecasting ability. The FA-LSTM model introduced in this study presents an effective new approach for daily runoff prediction.

A conditional latent autoregressive recurrent model for generation and forecasting of beam dynamics in particle accelerators

Particle accelerators are complex systems that focus, guide, and accelerate intense charged particle beams to high energy. Beam diagnostics present a challenging problem due to limited non-destructive measurements, computationally demanding simulations, and inherent uncertainties in the system. We propose a two-step unsupervised deep learning framework named as Conditional Latent Autoregressive Recurrent Model (CLARM) for learning the spatiotemporal dynamics of charged particles in accelerators. CLARM consists of a Conditional Variational Autoencoder transforming six-dimensional phase space into a lower-dimensional latent distribution and a Long Short-Term Memory network capturing temporal dynamics in an autoregressive manner. The CLARM can generate projections at various accelerator modules by sampling and decoding the latent space representation. The model also forecasts future states (downstream locations) of charged particles from past states (upstream locations). The results demonstrate that the generative and forecasting ability of the proposed approach is promising when tested against a variety of evaluation metrics.

Forecasting rheumatoid arthritis patient arrivals by including meteorological factors and air pollutants

The burden of rheumatoid arthritis (RA) has gradually elevated, increasing the need for medical resource redistribution. Forecasting RA patient arrivals can be helpful in managing medical resources. However, no relevant studies have been conducted yet. This study aims to construct a long short-term memory (LSTM) model, a deep learning model recently developed for novel data processing, to forecast RA patient arrivals considering meteorological factors and air pollutants and compares this model with traditional methods. Data on RA patients, meteorological factors and air pollutants from 2015 to 2022 were collected and normalized to construct moving average (MA)- and autoregressive (AR)-based and LSTM models. After data normalization, the root mean square error (RMSE) was adopted to evaluate models’ forecast ability. A total of 2422 individuals were enrolled. Not using the environmental data, the RMSEs of the MA- and AR-based models’ test sets are 0.131, 0.132, and 0.117 when the training set: test set ratio is 2:1, 3:1, and 7:1, while they are 0.110, 0.130, and 0.112 for the univariate LSTM models. Considering meteorological factors and air pollutants, the RMSEs of the MA- and AR-based model test sets were 0.142, 0.303, and 0.164 when the training set: test set ratio is 2:1, 3:1, and 7:1, while they were 0.108, 0.119, and 0.109 for the multivariable LSTM models. Our study demonstrated that LSTM models can forecast RA patient arrivals more accurately than MA- and AR-based models for datasets of all three sizes. Considering the meteorological factors and air pollutants can further improve the forecasting ability of the LSTM models. This novel method provides valuable information for medical management, the optimization of medical resource redistribution, and the alleviation of resource shortages.

Precipitation nowcasting method based on Spatial-Temporal Dual Discriminators

Abstract The prediction of radar echo sequence images is a spatiotemporal sequence forecasting problem, which is one of the main challenges in precipitation nowcasting. Addressing issues such as poor extraction of spatiotemporal features by previous models and blurry image sequence predictions, this study proposes a Spatial-Temporal dual Discriminator Precipitation Nowcasting Model (STD-SNGAN) based on spectral normalization generative adversarial networks (SNGAN). The model utilizes multi-scale convolution modules (Inception) to extract spatial features from radar echo images and convolutional gated recurrent units (ConvGRU) to extract temporal features between feature maps. By introducing a hidden feature sampler to enhance the extraction capability of the convolutional gated recurrent units and designing spatial-temporal dual discriminators to constrain the generator’s predicted samples, the spatiotemporal forecasting ability is enhanced. Experimental results demonstrate that the proposed STD-SNGAN model outperforms other algorithms in critical success index (CSI) and probability of detection (POD) for high echo intensity and long-term regions.

Does Machine Learning Forecast Investor’s Risk Appetite?

Risk appetite is an important indicator that is monitored with interest by financial market participants. One of the risk appetite indices is nominated “RISE risk appetite index” calculated to measure the riskiness of the Turkey market in general. There are very limited studies in the literature on RISE risk appetite, and most of them use simple econometric methods to predict the risk appetite. To the best of our knowledge, there is no study using machine learning algorithms. Therefore, it creates curiosity on how the success will be in estimating the risk appetite using machine learning algorithms. Thus, the aim of this paper is to measure the estimation success of the RISE index using Long Short-term Memory (LSTM) and Multi-Layer Perceptron (MLP). The analysis is based on a weekly frequency dataset covering the years 2008 to 2023. The results are compared as per RMSE values, and LSTM presents rather high prediction success compared to MLP. Due to the forecasting ability of BIST 100 index on RISE, the current and lagged values of BIST 100 are compared, and lagged values of BIST 100 are found to have a higher ability to estimate RISE, approximately twice as much as current values. It is expected that this valuable finding will be a guide for market participants and financial analysts to shape their investment preferences by using deep learning algorithms in predicting market expectations and to make well-directed investments.

Research on time series prediction of hybrid intelligent systems based on deep learning

Power forecasting plays a crucial role in the operation of smart grid system, which is indispensable for making the operation plan of power system, improving economic efficiency and ensuring the quality of power supply. In order to enhance the accuracy of power load forecasting, a hybrid intelligent power load forecasting system is proposed in this paper. The system first preprocesses the raw data using Savitzky-Golay smoothing technology to eliminate noise and improve data quality. Then, a long and short term memory network with attention mechanism is used to enhance the generalization ability of the model. In addition, in order to further improve the prediction performance, an improved genetic algorithm is integrated to optimize the model parameters. Finally, a data set is used to verify the proposed prediction method. In terms of short-term forecasting ability of experiment of the testing data set, compared with LSTM model, the proposed method shows superior performance in root mean square error and mean absolute error indicators, with root mean square error reduced by 18.7 % and mean absolute error reduced by 26.2 %. In terms of long-term prediction ability of experiment of the testing data set, compared with GBRT model, the proposed method reduces root mean square error and mean absolute error by 24.8 % and 30.7 %, respectively. The experimental results show that compared with the existing benchmark algorithm, the proposed method is significantly improved in two key indexes of prediction accuracy, which proves its effectiveness and superiority in power load prediction.

Does M&A activity spin the cycle of energy prices?

This research investigates the predictive power of mergers and acquisitions (M&A) activity on returns and volatility in energy commodities from January 1997 to September 2023. Utilizing a novel time-varying robust Granger causality framework, we analyse the dynamic relationship between M&A activity and energy returns and volatility within the global oil and gas (O&G) industry. In addition, we examine the network structure of M&A activity and energy prices across different quantile regimes. We find that M&A activity exhibits significant time-varying forecasting ability for both energy returns and volatility. Specifically, M&A transactions led by oil acquirers, representing deals where both the acquirer and target are within the O&G industry, demonstrate stronger forecasting ability for energy returns than M&A transactions led by acquirers from non-O&G industries. Conversely, M&A activity by non-O&G acquirers shows greater predictive ability for energy volatility. Robustness checks support our main findings. First, our multi-horizon model reveals significant bi-directional causality between M&A activity and energy series for 3 and 6-month forecasting horizons, which affirms a lasting influence on energy returns and volatility. Second, the strength of connectedness at extreme quantiles surpasses that at the median, with its magnitude increasing over the forecasting horizon. Third, our baseline results remain stable across varying rolling window sizes. These findings have important implications for policymakers and investors, suggesting that M&A activity within the O&G industry should be considered when making decisions in the energy market, as it plays a crucial role in predicting the dynamic direction of energy prices.

A New Index of Option Implied Absolute Deviation

ABSTRACTThis paper proposes a new index of forward looking absolute deviation extracted from option prices. The new index, named absolute deviation index (ADIX), is model‐free and easy to compute using at‐the‐money call and put option prices. It is shown that the spread between volatility index (VIX) and ADIX captures departures from normality in the risk‐neutral distribution and an empirical analysis using S&P 500 options data for the time period 1996–2021 reveals that the spread carries significant forecasting ability with respect to future equity returns at short to medium horizons. Portfolio strategies that use the spread as a predictor of S&P 500 returns outperform buy‐and‐hold strategies in an out‐of‐sample mean‐variance asset allocation exercise.

Spatiotemporal dynamics of epidemiology diseases: mobility based risk and short-term prediction modeling of COVID-19.

Nowadays, epidemiological modeling is applied to a wide range of diseases, communicable and non-communicable, namely AIDS, Ebola, influenza, Dengue, Malaria, Zika. More recently, in the context of the last pandemic declared by the World Health Organization (WHO), several studies applied these models to SARS-CoV-2. Despite the increasing number of researches using spatial analysis, some constraints persist that prevent more complex modeling such as capturing local epidemiological dynamics or capturing the real patterns and dynamics. For example, the unavailability of: (i) epidemiological information such as the frequency with which it is made available; (ii) sociodemographic and environmental factors (e.g., population density and population mobility) at a finer scale which influence the evolution patterns of infectious diseases; or (iii) the number of cases information that is also very dependent on the degree of testing performed, often with severe territorial disparities and influenced by context factors. Moreover, the delay in case reporting and the lack of quality control in epidemiological information is responsible for biases in the data that lead to many results obtained being subject to the ecological fallacy, making it difficult to identify causal relationships. Other important methodological limitations are the control of spatiotemporal dependence, management of non-linearity, ergodicy, among others, which can impute inconsistencies to the results. In addition to these issues, social contact, is still difficult to quantify in order to be incorporated into modeling processes. This study aims to explore a modeling framework that can overcome some of these modeling methodological limitations to allow more accurate modeling of epidemiological diseases. Based on Geographic Information Systems (GIS) and spatial analysis, our model is developed to identify group of municipalities where population density (vulnerability) has a stronger relationship with incidence (hazard) and commuting movements (exposure). Specifically, our framework shows how to operate a model over data with no clear trend or seasonal pattern which is suitable for a short-term predicting (i.e., forecasting) of cases based on few determinants. Our tested models provide a good alternative for when explanatory data is few and the time component is not available, once they have shown a good fit and good short-term forecast ability.

Short‐Term Sea Fog Area Forecast: A New Data Set and Deep Learning Approach

AbstractPrompt and precise forecast of sea fog regions ensures maritime navigational safety. This paper establishes a prompt and precise forecast of sea fog regions that ensure maritime navigational safety. This paper establishes a multivariable sea fog forecast (MV‐SFF) data set and proposes a deep learning‐based forecast method named rich‐element aggregated (REA) for short‐term forecasts. The MV‐SFF data set contains 122 sea fog events from 2010 to 2020. Each event in the MV‐SFF data set includes meteorological variables from reanalysis data and geostationary ocean color imager satellite images captured hourly on the day of the sea fog occurrence. This data set aims to comprehensively utilize meteorological elements and remote sensing observations to predict the spatial changes of sea fog areas in the next 7 hours. The proposed REA model can extract and integrate features from historical meteorological elements of different types, times, and spatial locations. In addition, REA utilizes a position‐aware edge detection mechanism to locate the exact position of sea fog. We perform a seven‐hour ahead forecast starting from 09:16 local time and show promising forecasting results, which demonstrate the fog area forecast ability of our model. Compared with existing advanced deep learning networks, the REA model is superior in performance and stability.

<span>Predicting learning status of informatics students at Dong Thap University based on machine learning models</span>

For effective student support, it is essential to forecast student academic performance status. This study explores student academic performance and applies four machine learning (ML) models to predict student learning alert status. The research collects input data to build a model including parameters such as entrance grades, information about accommodation, learning environment, and first-semester result from undergraduates of the Faculty of Mathematics and Computer Education, Dong Thap University. The effective ML techniques such as Logistic regression (LR), Support vector machine (SVM), Decision trees (DTs), and Random Forest (RF) are suggested to predict student learning alert status. The outcomes are evaluated on metrics like Accuracy, Precision, Recall, and F1 score. The results show that the forecasting ability of logistic regression models outperforms in classifying student performance compared to other methods by yielding optimal classification results like high accuracy and Sensitivity followed by RF, SVM, and DTs.

Can U.S. macroeconomic indicators forecast cryptocurrency volatility?

This research examines the ability of 28 U.S. macroeconomic variables to forecast the volatility of six cryptocurrencies. In- and out-of-sample analyses are performed to validate their forecasting ability. Our analysis shows that during the full-sample period, 15 variables display forecasting ability, while post-COVID-19 period, this number is 17. Among these variables, the most influential include the consumer confidence index, leading economic index, consumer price index, U.S. exports and U.S. imports. Importantly, the predictive ability of these variables appears to have strengthened during the post-COVID-19 period. The out-of-sample results confirm the effectiveness of those macroeconomic variables in the in-sample tests. Furthermore, the robustness test reveals that incorporating these U.S. macroeconomic variables can enhance the performance of the GARCH volatility model. In this study, combination methods are used to enhance forecasting stability and are proven to have good forecasting ability. Our research also indicates that integrating global macroeconomic variables can enhance forecasting ability while recognizing the valuable information provided by U.S. macroeconomic variables. Additionally, we find that variables such as the short-term government bond yield and the M1 money supply emerge as important predictors of cryptocurrency bubbles.