JournalTOCs

About Help News

Turkish Journal of Forecasting
Number of Followers: 1

This is an Open Access Journal

Open Access journal
ISSN (Online) 2618-6594
Published by DePauw University

[3 journals]

Impact of Structural Break Location on Forecasting Accuracy: Traditional
Methods Versus Artificial Neural Network
- Authors: Daud Ali ASER; Esin FİRUZAN
  Abstract: Since forecasting future values is fundamental for researchers, investors, practitioners, etc., obtaining accurate predictions is critical in time series analysis. The accuracy is reliant on good modelling and good-quality data. The latter is affected by unusual observations, changes over time, missing data, and structural breaks among others. Economic crises are the major cause of data instability and therefore, this paper focuses on how structural breaks in conditional heteroscedastic financial and macroeconomic data affect forecasting accuracy on short and long-term horizons. More specifically, we are interested in the impact of the location of the structural break and break size on the predictive performance of two linear (ARIMA and Exponential Smoothing) forecasting models and two nonlinear (ARIMA – ARCH and Artificial Neural Network) models. We conducted Monte Carlo simulations and showed that the forecasting accuracy decreases as the structural break location approaches the end of the sample. In addition, break size and length of the horizon show the same impact on the forecasting accuracy as the forecasting error increases with the increase of break magnitude and length of the horizon. We also showed that ARIMA – ARCH model is the best performing in the absence of a structural break while the artificial neural network model outperforms all the competing models in the presence of structural break, especially in large break sizes and long horizons. Last, we applied the above techniques to forecasting daily close prices of brent oil and Turkish Lira – USD exchange rates out–of–sample, and similar results were found.
  PubDate: Sat, 31 Dec 2022 00:00:00 +030
Topic Modelling and Artificial Intelligence based Method Using Online
Employee Assessments to Analyse Job Satisfaction
- Authors: Ali ÖZDEMİR; Aytuğ ONAN, Vildan ÇINARLI ERGENE
  Abstract: In this study, the performance of the proposed sample selection method was evaluated on some basic classifiers by conducting a basic literature review on the use of topic modelling methods by considering the online evaluations of the employees in order to determine and analyse the job satisfaction factors. In addition, the effectiveness of different representation structures is evaluated in order to represent the data sets effectively and the main results are obtained regarding the use of classification ensemble methods in the field of text mining. In this work it was emphasized that machine learning methods can achieve high performance in classification and work effectively and scalable with large data sets. The dataset used in this study was obtained from www.kaggle.com. A total of 67529 comments collected from people working at Google, Amazon, Netflix, Facebook, Apple, and Microsoft were evaluated. Within the scope of this study, a text mining and artificial intelligence-based method will be developed, and a solution will be brought to text mining with artificial intelligence methods.
  PubDate: Sat, 31 Dec 2022 00:00:00 +030
The Effect of Handling Imbalanced Datasets Methods on Prediction of
Entrepreneurial Competency in University Students
- Authors: Murat ŞİMŞEK; Ahmet Said DAŞ
  Abstract: As of today, entrepreneurs and entrepreneurship are considered to be the integral parts of the economic and technological advancements. Entrepreneurs are promoted in many countries because of their high return on investment opportunities both in terms of income and new inventions. Numerous studies prove that entrepreneurs have many traits in common and these common traits can correlate with each other. Based on these common traits, potential entrepreneurs can be predicted, current entrepreneurs can be improved by realising their weak sides and the ones who wish to be entrepreneurs can be provided with insights. A machine learning approach can light the way for a better rewarding future for entrepreneurship, helping these goals significantly. There exist several studies for the prediction of entrepreneurial competency with the use of machine learning algorithms. Most machine learning methods perform better accuracy and F1-score imbalanced data instead in imbalanced data. This study focuses on utilizing imbalanced class handling methods to increase prediction performance. Random Oversampling, Random Undersampling, SMOTE, and NearMiss methods are used to handling imbalanced data for this purpose in this study. The performance of the machine learning algorithms with Imbalanced Data Handling methods is compared with the machine learning algorithms without these methods. The comparison shows that with the handling imbalanced data methods machine learning algorithms perform better.
  PubDate: Sat, 31 Dec 2022 00:00:00 +030
Convolutional Neural Networks for MRI-Based Brain Tumor Segmentation: A
Comparative Analysis of State-of-the-Art Segmentation Networks
- Authors: Ahmet Furkan BAYRAM; Caglar GURKAN, Abdulkadir BUDAK, Hakan KARATAŞ
  Abstract: The prevalence of brain tumor is quite high. Brain tumor causes critical diseases. Also, brain tumor causes a variety of symptoms in most people. This study aims to segmentation of the tumor in the brain. For this purpose, state-of-art architectures, such as UNet, Attention UNet, Residual UNet, Attention Residual UNet, Residual UNet++, Inception UNet, LinkNet, and SegNet were used for segmentation. 592 magnetic resonance (MR) images were utilized in the training and testing of segmentation architectures. In the comparative analysis, Attention UNet achieved the best predictive performance with a 0.886 dice score, 0.795 IoU score, 0.881 sensitivity, 0.993 specificity, 0.891 precision, and 0.986 accuracy.
  PubDate: Sat, 31 Dec 2022 00:00:00 +030
Fully Automatic End-to-End Convolutional Neural Networks-Based Pancreatic
Tumor Segmentation on CT Modality
- Authors: Ahmet Furkan BAYRAM; Caglar GURKAN, Abdulkadir BUDAK, Hakan KARATAŞ
  Abstract: The pancreas is one of the vital organs in the human body. Early diagnosis of a disease in the pancreas is critical. In this way, the effects of pancreas diseases, especially pancreatic cancer on the person are decreased. With this purpose, artificial intelligence-assisted pancreatic cancer segmentation was performed for early diagnosis in this paper. For this aim, several state-of-the-art segmentation networks, UNet, LinkNet, SegNet, SQ-Net, DABNet, EDANet, and ESNet were used in this study. In the comparative analysis, the best segmentation performance has been achieved by SQ-Net. SQ-Net has achieved a 0.917 dice score, 0.847 IoU score, 0.920 sensitivity, 1.000 specificity, 0.914 precision, and 0.999 accuracy. Considering these results, an artificial intelligence-based decision support system was created in the study.
  PubDate: Sat, 31 Dec 2022 00:00:00 +030
Prediction of the Premium Production of Some Insurance Companies Operating
in Turkey with Artificial Neural Networks
- Authors: Buse ÖZGÜR; Ufuk YOLCU
  Abstract: The insurance sector can be seen as a sector that directly affects the country's economy and development with its ability to fund financial markets and meet risks. In this respect, estimating the premium size, which is the main factor that constitutes the volume of the insurance sector, as accurately and reliably as possible, indirectly means predicting the risks that may arise in terms of the economy and development of the country and taking precautions. necessary measures. In this study, premium productions of some insurance companies operating in Turkey were estimated with different artificial neural networks and their results were evaluated comparatively. In this context, two different artificial neural networks (ANNs), feed forward and feedback, were used as the estimation tools for insurance premium production. Two training algorithms and two different activation functions were run in the structure of the ANNs used. Thus, eight different estimation tools were created for insurance companies' premium production. The estimation performances of ANNs were evaluated on test sets by using error criteria such as Root Mean Square Error, Mean Absolute Percentage Error, and Median Absolute Percentage Error. In terms of the MdAPE criterion in our best-performing algorithms, in the analysis of a total of 36 data sets, 18 quarters of 18 months in total, the predictions for only 6 data sets were estimated with an error of more than 10%, and 5 of them were around 10% or just above, which is still acceptable. have an acceptable level of error.
  PubDate: Sat, 31 Dec 2022 00:00:00 +030

JournalTOCs
School of Mathematical and Computer Sciences
Heriot-Watt University
Edinburgh, EH14 4AS, UK
Email: journaltocs@hw.ac.uk
Tel: +00 44 (0)131 4513762

Your IP address: 3.237.29.69

Home (Search)
API
About JournalTOCs
News (blog, publications)