Search

In this study, three new regression models are created for magnitude-type conversion with different machine learning algorithms (linear regression, regression trees, support vector machines, Gaussian process regression models, ensembles of trees) by using the earthquakes (M ≥ 4.0) that occurred in Turkey (1900–2020). Additionally, eight new equations are formed with linear and orthogonal regression methods. Developed equations and models are compared to equations selected from the literature by test data. As a result of the study, it is observed that machine learning algorithms create better models and provide results closer to the real values than created and selected equations. [ABSTRACT FROM AUTHOR]

In Soma Region, located in the western part of Turkey, both tectonic and anthropogenic source type of events are identified in earthquake catalogues by various seismology centers. Especially because of the low-magnitude quarry blasts, mistakes can be made in seismic catalogs in terms of source type, which can be further complicated by similar magnitude earthquakes. With this regard, the vertical component seismograms of 445 seismic events with a magnitude of Ml ≤ 2.5 recorded at the station SOMA, operated by Boğaziçi University Kandilli Observatory and Earthquake Research Institute Regional Earthquake–Tsunami Monitoring Center (KOERI–RETMC), were analyzed. First, two statistical analyses (Linear and Quadratic Discriminant Functions–LDF/QDF) were applied for amplitude ratio and complexity methods for 345 waveforms that have the same source types in both KOERI–RETMC catalogs and the first manual determination. The accuracies of the statistical approaches are varied between 87.25% and 97.39% and the better statistical classifier is the QDF for complexity method. Then, using the values obtained from two methods together, tree-based machine learning (ML) classifiers called as Random Forest (RF), Gradient Boosting (GB) and Support Vector Machine (SVM) were applied to the same data set. All classifiers provided as 100% success rate for quarry blasts' recordings, while earthquakes are categorized by RF, GB and SVM with 97.1%, 95.8% and 92.8%, respectively. Each ML algorithms were applied to the other 100 data identified as quarry blast on KOERI–RETMC catalogs but determined to be suspicious source types on first manual determination. Regarding to the outperforming RF and GB algorithms, the quarry blast recordings have just been reached as 53 and 55, respectively. Considering the accuracies of the ML algorithms in the testing and training data set, the source types of the low magnitude seismic events that are registered in the catalogs should be re-evaluated and refined in Soma Region using the station SOMA. [ABSTRACT FROM AUTHOR]

Climate plays a dominant role in influencing the process of evaporation and is projected to have adverse effects on water resources especially in the wake of a changing climate. In order to understand the impact of climate change on water resources, artificial intelligence models that possesses rapid decision-making ability, are used. This study was carried out to estimate evaporation in the Karaidemir Reservoir in Turkey with artificial neural networks (ANNs). The daily meteorological data covering the irrigation season were provided for a 30-year reference period and used to develop artificial neural network models. Predicted meteorological data based on climate change projections of HadGEM2-ES and MPI-ESM-MR under the Representative Concentration Pathway (RCP) 4.5 and 8.5 future emissions scenarios between 2000–2098 were utilized for future evaporation projections. The study also focuses on optimal crop patterns and water requirement planning in the future. ANNs model was run for each of the scenarios created based on ReliefF algorithm results using different testing-training-validation rates and learning algorithms of Bayesian Regularization (BR), Levenberg–Marquardt (L-M) and Scaled Conjugate Gradient (SCG). The performance of each alternative model was compared with coefficient of determination (R2) and mean square error (MSE) measures. The obtained results revealed that the ANNs model has high performance in estimation with a few input parameters, statistically. Projected surface water evaporation for the long term (2080–2098) showed an increase of 1.0 and 3.1% for the RCP4.5 scenarios of the MPI and HadGEM model, and a 14% decrease and 7.3% increase for the RCP8.5 scenarios, respectively. [ABSTRACT FROM AUTHOR]

Home healthcare services are public or private service that aims to provide health services at home to socially disadvantaged, sick, needy, disabled, and elderly individuals. This study aims to increase the quality of home healthcare practice by analyzing the factors affecting it. In Megacity Istanbul, data from 1707 patients were used by considering 14 different input variables affecting home healthcare practice. The demographic, geographic, and living conditions of patients and healthcare professionals who take an active role in home healthcare practice constituted the central theme of the input parameters of this study. The regression method was used to look at the factors that affect the length of time a patient needs home healthcare, which is the study's output variable. This article provides short planning times and flexible solutions for home healthcare practice by showing how to avoid planning patient healthcare applications by hand using methods that were developed for home health services. In addition, in this research, the AB, RF, GB, and NN algorithms, which are among the machine learning algorithms, were developed using patient and personnel data with known input parameters to make home healthcare application planning correct. These algorithms' accuracy and error margins were calculated, and the algorithms' results were compared. For the prediction data, the AB model showed the best performance, and the R2 value of this algorithm was computed as 0.903. The margins of error for this algorithm were found to be 0.136, 0.018, and 0.043 for the RMSE, MSE, and MAE, respectively. This article provides short planning times and flexible solutions in home healthcare practice by avoiding manual patient healthcare application planning with the methods developed in the context of home health services. [ABSTRACT FROM AUTHOR]

The study aimed to describe and test machine learning (ML)‐based algorithms to evaluate the unit price of drinking milk. The algorithms were applied to the data collected over 8 years in 2014 and 2021 related to the price of drinking milk in Turkey. The economic, social, and environmental factors that have an impact on the unit price of drinking milk were evaluated. Five ML algorithms, including random forest, gradient boosting, support vector machine (SVM), neural network, and AdaBoost algorithms, were utilized to predict the drinking milk unit price. ML also applied hyperparameter tuning with nested cross‐validation to calculate the prediction accuracy for each algorithm. The results show that the random forest algorithm based on the features of the ML algorithms has the best performance, with the accuracy of 99.30% for training and 98.10% for testing the dataset. The average accuracy of gradient boosting, SVM, neural network, and AdaBoost are obtained as 97.30%, 96.15%, 95.65%, and 96.05%, respectively. Random forest performed best as the target variable with the lowest deviation values of mean squared error (MSE) (0.004), root mean square error (RMSE) (0.060), and mean absolute error (MAE) (0.029) in the training and MSE (0.009), RMSE (0.096), and MA (0.055) in the testing dataset. This study presents an interesting perspective with practical potential to adopt ML methods in the dairy industry. The developed ML algorithms can provide dairy investors and policymakers with important decision‐support information. [EconLit Citations: C13, C53, L66, C88]. [ABSTRACT FROM AUTHOR]

Purpose: This paper aims to determine priority issues in the corporate governance (CG) principles to increase CG rating notes of publicly traded companies. Design/methodology/approach: This study defines the priority issues for publicly traded companies that should be focused to increase the CG rating notes. In this context, this study considers the companies in Borsa Istanbul CG index (XKURY), use data for 2018, 2019, 2020, and applies machine learning algorithms. Findings: Overall, importance of each CG principle changes for the CG rating notes; first five CG principles in terms of significance have a total of 43.6% importance for the CG rating notes; following a straight-line approach in completing deficiencies of the CG principles cannot help increase the CG rating notes. Hence, empirical results highlight the impact of the most significant CG principles in terms of the CG rating notes that should be focused on by publicly traded companies so that CG ratings can be increased. Research limitations/implications: This study uses Turkey data and considers publicly traded companies in the XKURY index. The main cause of this condition is that consolidated data of compliance report format for all publicly traded companies cannot be obtained. Practical implications: The publicly traded companies can increase the CG rating notes by considering the results of this study while focusing on priority issues in the CG principles. Social implications: The study determines the most important CG principles that companies can focus on, highlights the importance of usage of machine learning algorithms in determining the most influential CG principles in terms of the CG rating notes and reflects on the difficulties for gathering consolidated CG principles compliance reporting data for all publicly traded companies. Hence, societies can have better companies that are ruled more efficiently and corporately by increasing their compliance with the CG principles. Originality/value: To the best of the authors' knowledge, this is the first empirical study that determines the priority issues to increase the CG rating notes of publicly traded companies based on the new CG principles compliance reporting scheme in Turkey. Following this aim, machine learning algorithms, which can present better results with regard to most of the econometric models, are used in this study. [ABSTRACT FROM AUTHOR]

This research aims to estimate the delivery time and energy cost of e-scooter vehicles for distributing mail or packages and to show the usage efficiency of e-scooter sharing services in postal service delivery in Turkey. The machine learning (ML) methods used to implement the prediction of delivery time and energy cost as output variables include random forest (RF), gradient boosting (GB), k-nearest neighbour (kNN), and neural network (NN) algorithms. Fifteen input variables under demographic, environmental, geographical, time, and meta-features are utilised in the ML algorithms. The correlation coefficient (R2) values of RF, GB, NN, and kNN algorithms were computed for delivery time as 0.816, 0.845, 0.821, and 0.786, respectively. The GB algorithm, which has a high R2 and the slightest margin of error, exhibited the best prediction performance for delivery time and energy cost. Regarding delivery time, the GB algorithm's MSE, RMSE, and MAE values were calculated as 149.32, 12.22, and 6.08, respectively. The R2 values of RF, GB, NN, and kNN algorithms were computed for energy cost as 0.917, 0.953, 0.400, and 0.365, respectively. The MSE, RMSE, and MAE values of the GB algorithm were calculated as 0.001, 0.019, and 0.009, respectively. The average energy cost to complete a package or mail delivery process with e-scooter vehicles is calculated as 0.125 TL, and the required time is approximately computed as 11.21 min. The scientific innovation of the study shows that e-scooter delivery vehicles are better for the environment, cost, and energy than traditional delivery vehicles. At the same time, using e-scooters as the preferred way to deliver packages or mail has shown how well the delivery service works. Because of this, the results of this study will help in the development of ways to make the use of e-scooters in delivery service even more efficient. [ABSTRACT FROM AUTHOR]

The study examines the role of data frequency and estimation methods in electricity price estimation by applying selected machine learning algorithms and time series econometric models. In this context, Turkey is selected as an emerging country example, seven explanatory variables including COVID-19 pandemic is considered, and daily and weekly data between February 20, 2019 and March 26, 2021 that includes pre-pandemic and pandemic periods are used. The empirical results show that (i) machine learning algorithms perform better than time series econometric models for both pre-pandemic and pandemic periods; (ii) high-frequency data increases the performance of estimation models; (iii) machine learning algorithms perform better with high-frequency (daily) data with regard to low-frequency (weekly) data; (iv) the pandemic causes an adverse effect on the performance of estimation models; (v) energy-related variables are more important than other variables although all are significant; (vi) the share of renewable sources in electricity production is the most important variable on the electricity prices in both periods and data types. Hence, the findings highlight the role of data frequency and method selection in electricity prices estimation. Moreover, policy implications are discussed. • The study examines role of data frequency and prediction methods in electricity prices. • This study applies machine learning algorithms and time series econometric models. • Machine learning algorithms perform better than time series econometric models. • High-frequency data increases performance of prediction models. • COVID-19 pandemic causes an adverse effect on performance of prediction models. [ABSTRACT FROM AUTHOR]

The study aims to define the sources of Turkey’s sovereign CDS spread changes to develop policies that stabilize CDS spreads since they have a volatile and increasing trend, especially in the last two years. In this context, monthly data of 13 factors related to international, macroeconomic, and market between 2011/1 and 2019/12 are used by dividing the dataset into three periods as the full period (2011-2019), the stability period (2011-2017), and the macroeconomic turbulent period (2018-2019) and performing 4 different machine learning algorithms. The empirical results prove that (i) Treasury bond interest rate should be lower than 8% in the stability period and gold prices should be lower than TL 5.500 in the macroeconomic turbulent period to have low-level CDS spreads; (ii) NPL volume has no significant effect on in any period examined; (iii) the significance of factors on sovereign CDS spreads vary over the periods. [ABSTRACT FROM AUTHOR]

Copyright of Istanbul Commerce University Journal of Social Sciences / İstanbul Ticaret Üniversitesi Sosyal Bilimler Dergisi is the property of Istanbul Commerce University Journal of Social Sciences and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Some countries have announced national benchmark rates, while others have been working on the recent trend in which the London Interbank Offered Rate will be retired at the end of 2021. Considering that Turkey announced the Turkish Lira Overnight Reference Interest Rate (TLREF), this study examines the determinants of TLREF. In this context, three global determinants, five country-level macroeconomic determinants, and the COVID-19 pandemic are considered by using daily data between December 28, 2018, and December 31, 2020, by performing machine learning algorithms and Ordinary Least Square. The empirical results show that (1) the most significant determinant is the amount of securities bought by Central Banks; (2) country-level macroeconomic factors have a higher impact whereas global factors are less important, and the pandemic does not have a significant effect; (3) Random Forest is the most accurate prediction model. Taking action by considering the study's findings can help support economic growth by achieving low-level benchmark rates. [ABSTRACT FROM AUTHOR]

The motion picture industry is one of the largest industries worldwide and has significant importance in the global economy. Considering the high stakes and high risks in the industry, forecast models and decision support systems are gaining importance. Several attempts have been made to estimate the theatrical performance of a movie before or at the early stages of its release. Nevertheless, these models are mostly used for predicting domestic performances and the industry still struggles to predict box office performances in overseas markets. In this study, the aim is to design a forecast model using different machine learning algorithms to estimate the theatrical success of US movies in Turkey. From various sources, a dataset of 1559 movies is constructed. Firstly, independent variables are grouped as pre-release, distributor type, and international distribution based on their characteristic. The number of attendances is discretized into three classes. Four popular machine learning algorithms, artificial neural networks, decision tree regression and gradient boosting tree and random forest are employed, and the impact of each group is observed by compared by the performance models. Then the number of target classes is increased into five and eight and results are compared with the previously developed models in the literature. [ABSTRACT FROM AUTHOR]

• The direction of the financial dollarization rate is predicted one week ahead. • Four machine learning algorithms are utilized. • Parameters of the machine learning algorithms are optimized with genetic algorithm. Financial dollarization has many implications on the economy and the banking sector of developing countries. High level of financial dollarization causes fragilities on the balance sheet of the banks. Due to the negative effects of dollarization on the banks, predicting the financial dollarization rate for the following periods becomes very crucial for the soundness of the banking sector. This study aims to predict the weekly movement of the financial dollarization rate in the Turkish banking sector. The dataset includes asset dollarization rate series for 848 weeks between 30/12/2005 and 25/03/2022. Four different machine learning algorithms (K-Nearest Neighbor, Decision Tree, Naïve Bayes, and Support Vector Machine) are utilized to predict the next week's financial dollarization rate movement. The parameters of the machine learning algorithms are optimized with a genetic algorithm. The parameters are divided into two groups as common parameters and model-specific parameters. Common parameters are the parameters utilized in all machine learning models and include the score transform method, standardization choice, the values on the cost matrix (used to reduce the misclassification rate), and autoregressive degree. The overall dataset is divided into four sub-periods and three different predicting schemes are utilized in each sub-period and overall period. According to the results of the analysis, the prediction performance in the overall dataset, which covers a wider period, was up to 73 % and the prediction performance was up to 90 % in sub-period datasets where the economy was relatively stable. [ABSTRACT FROM AUTHOR]

• Prediction accuracy of clay was more affected by sampling strategies. • Vis-NIR spectra had higher prediction performance compared to pXRF spectra. • Combined Vis-NIR and pXRF spectra had no improvement on prediction accuracy. The recent technologies employed for rapid, cost-effective, and non-destructive prediction of soil particle size distribution (clay, sand, and silt) are becoming increasingly interesting among soil scientists. Our aims were to explore the effect of surface, profile wall, and surface + profile wall on prediction accuracy using individual and combined both soil spectra (Vis-NIR and pXRF) with machine learning algorithms for sand, silt, and clay. In total, 191 soil samples were collected from the soil surface (0–30 cm) and profile wall (1 m × 1 m) from cultivated fields in Eskisehir, Central Anatolia of Türkiye. The pXRF (0–45 keV) and Vis-NIR (350–2500 nm) spectroradiometers were used to obtain soil spectra from sieved soil samples. The prediction accuracy of each soil particle size was evaluated by 54 models to explore the predictive performance. The five machine learning algorithms (elastic net, lasso, random forest, ridge, and support vector machine-linear) were applied with calibration (70% soil samples) and validation (30% soil samples) data set for each soil particle size.Results showed the dominant clay mineral in the A and C horizons is chlorite. Moderate and high prediction accuracy for sand (R2 = 0.56–0.84) and clay (R2 = 0.61–0.80), whereas only moderate prediction accuracy for silt (R2 = 0.47–0.55) using both soil spectra in the surface, profile wall, and surface + profile wall. The highest prediction accuracy for each soil particle size was achieved in the soil profile wall using Vis-NIR spectra with elastic net, which outperformed other samplings such as individual pXRF, combined both soil spectra, and other machine learning algorithms. In addition, the prediction accuracy of clay was more affected by sampling strategies compared to sand and silt. We concluded that individual Vis-NIR spectroradiometer can be utilized to achieve the highest prediction accuracy for sand, silt, and clay ratio in semiarid ecosystems for soil surveys and land use studies. [ABSTRACT FROM AUTHOR]

The prediction of global solar radiation for the regions is of great importance in terms of giving directions of solar energy conversion systems (design, modeling, and operation), selection of proper regions, and even future investment policies of the decision-makers. With this viewpoint, the objective of this paper is to predict daily global solar radiation data of four provinces (Kırklareli, Tokat, Nevşehir and Karaman) which have different solar radiation distribution in Turkey. In the study, four different machine learning algorithms (support vector machine (SVM), artificial neural network (ANN), kernel and nearest-neighbor (k-NN), and deep learning (DL)) are used. In the training of these algorithms, daily minimum and maximum ambient temperature, cloud cover, daily extraterrestrial solar radiation, day length and solar radiation of these provinces are used. The data is supplied from the Turkish State Meteorological Service and cover the last two years (01.01.2018–31.12.2019). To decide on the success of these algorithms, seven different statistical metrics (R2, RMSE, rRMSE, MBE, MABE, t-stat, and MAPE) are discussed in the study. The results shows that R2, MABE, and RMSE values of all algorithms are ranging from 0.855 to 0.936, from 1.870 to 2.328 MJ/m2, from 2.273 to 2.820 MJ/m2, respectively. At all cases, k-NN exhibited the worst result in terms of R2, RMSE, and MABE metrics. Of all the models, DL was the only model that exceeded the t-critic value. In conclusion, the present paper is reporting that all machine learning algorithms tested in this study can be used in the prediction of daily global solar radiation data with a high accuracy; however, the ANN algorithm is the best fitting algorithm among all algorithms. Then it is followed by DL, SVM and k-NN, respectively. • Comparison of four machine learning algorithms in the prediction of daily global solar radiation. • Discussion of the performance success of the algorithms with seven metrics. • Different regions come different algorithms to the fore in terms of prediction success. • ANN is generally presenting better prediction results in comparison with those of DL, SVM, and k-NN. • k-NN algorithm is giving the worst prediction results for almost all metrics. [ABSTRACT FROM AUTHOR]