Amaç- Büyük veriyi işleme noktasında şirketler klasik istatistik yöntemlerinden daha hızlı sonuca ulaşabilecekleri yapay zeka yöntemlerini
tercih etmeye başlamışlardır. Bu nedenle son yıllarda yapay zeka yöntemleriyle finansal performans tahmini çalışmaları da giderek
artmaktadır. Farklı zaman serisi modelleri ile finansal performans tahmini çalışmaları olmasına rağmen Facebook Prophet modeli ile yapılan
çalışmaya literatürde rastlanmadığından alandaki boşluğa katkı sağlanması amaçlanmaktadır.
Yöntem- BIST İmalat Sektöründe yer alan 173 şirketin Kamuyu Aydınlatma Platformu veri tabanından elde edilen 2009-2020 yılları arasındaki
verileri, net dönem kar/zararı hedef değişken olarak seçilerek yapay öğrenme modeli Prophet ile Python programı üzerinde çalışarak finansal
performans tahmini yapılmıştır.
Bulgular- Veri setindeki 46 dönem eğitildikten sonra son 2 döneminin tahminine ait MSE değerleri 0,0185 ile 25,0147 arasında, RMSE
değerleri 0,1361 ile 5,0015 ve MAPE değerleri ise 0,1002 ile 4,6634 aralığında ölçülmüştür. Sıfıra yakın değerler yanında sıfırdan uzaklaşan
değerler de bulunmaktadır. Bu da başarılı tahminlerin yanında başarısız tahminlerin de olduğunu göstermektedir.
Sonuç- Prophet yönteminin çok fazla çaba harcamayı gerektirmeden şirketlere zaman kazandıracağı söylenebilir. Doğruluğu ve performansı
daha da geliştirmek için birden fazla modelin güçlü yönlerinden yararlanarak karma bir yapay zeka modeli oluşturmak sonraki çalışmalar için
önerilmektedir.
Purpose- Through processing big data, companies have started to prefer artificial intelligence methods that can reach results faster than
classical statistical methods. For this reason, financial performance forecasting studies with artificial intelligence methods have been
increasing in recent years. The purpose of this study is to contribute to the gap in the field of financial performance forecasting with the
Facebook Prophet model which has not been applied before although there are other time series models in the literature.
Methodology- The study employs Facebook Prophet artificial learning model by choosing the net profit/loss as target variable for financial
performance forecasting on Python program with the data of 173 companies in the BIST Manufacturing Sector between 2009 and 2020,
obtained from the Public Disclosure Platform.
Findings- After training the 46 periods of the data set, performance matrices of MSE, RMSE and MAPE were measured for the last 2 periods
as MSE values between 0.0185 and 25.0147, RMSE values between 0.1361 and 5.0015 and MAPE values between 0.1002 and 4.6634. In
addition to values close to zero, there are also values that move away from zero. The analysis reveals that besides successful predictions,
there are also unsuccessful predictions.
Conclusion- It may be concluded that the Facebook Prophet method will save companies time without requiring much effort. To further
improve accuracy and performance, creating a mixed artificial intelligence model by leveraging the strengths of multiple models is
recommended for further studies.
financial performance forecasting artificial intelligence time series Facebook Prophet BIST manufacturing sector
Primary Language | Turkish |
---|---|
Subjects | Economics, Finance, Business Administration |
Journal Section | Articles |
Authors | |
Publication Date | December 31, 2021 |
Published in Issue | Year 2021 Volume: 8 Issue: 4 |
Journal of Economics, Finance and Accounting (JEFA) is a scientific, academic, double blind peer-reviewed, quarterly and open-access online journal. The journal publishes four issues a year. The issuing months are March, June, September and December. The publication languages of the Journal are English and Turkish. JEFA aims to provide a research source for all practitioners, policy makers, professionals and researchers working in the area of economics, finance, accounting and auditing. The editor in chief of JEFA invites all manuscripts that cover theoretical and/or applied researches on topics related to the interest areas of the Journal. JEFA publishes academic research studies only. JEFA charges no submission or publication fee.
Ethics Policy - JEFA applies the standards of Committee on Publication Ethics (COPE). JEFA is committed to the academic community ensuring ethics and quality of manuscripts in publications. Plagiarism is strictly forbidden and the manuscripts found to be plagiarized will not be accepted or if published will be removed from the publication. Authors must certify that their manuscripts are their original work. Plagiarism, duplicate, data fabrication and redundant publications are forbidden. The manuscripts are subject to plagiarism check by iThenticate or similar. All manuscript submissions must provide a similarity report (up to 15% excluding quotes, bibliography, abstract and method).
Open Access - All research articles published in PressAcademia Journals are fully open access; immediately freely available to read, download and share. Articles are published under the terms of a Creative Commons license which permits use, distribution and reproduction in any medium, provided the original work is properly cited. Open access is a property of individual works, not necessarily journals or publishers. Community standards, rather than copyright law, will continue to provide the mechanism for enforcement of proper attribution and responsible use of the published work, as they do now.