مدل بندي و تعيين توان مديريت سرمايه در گردش در پيش بيني ورشكستگي مالي شركت ها با استفاده از الگوريتم هاي هوش مصنوعي

Modeling and Determining the Power of Working Capital Management in Predicting Corporate Financial Bankruptcy Using Artificial Intelligence Algorithms

ﻫﺪف اﺻﻠﯽ اﯾﻦ ﭘﮋوﻫﺶ ﻣﺪلﺑﻨﺪي و ﺗﻌﯿﯿﻦ ﺗﻮان ﻣﺪﯾﺮﯾﺖ ﺳﺮﻣﺎﯾﻪ در ﮔﺮدش در ﭘﯿﺶﺑﯿﻨﯽ ورﺷﮑﺴﺘﮕﯽ ﻣﺎﻟﯽ ﺷﺮﮐﺖﻫﺎ ﺑﺎ اﺳﺘﻔﺎده از اﻟﮕﻮرﯾﺘﻢﻫﺎي ﻫﻮش ﻣﺼﻨﻮﻋﯽ اﺳﺖ. ﺟﺎﻣﻌﻪ آﻣﺎري ﭘﮋوﻫﺶ ﻣﺘﺸﮑﻞ از 120 ﺷﺮﮐﺖ ﭘﺬﯾﺮﻓﺘﻪ ﺷﺪه در ﺑﻮرس اوراق ﺑﻬﺎدار ﺗﻬﺮان ﻃﯽ ﺳﺎلﻫﺎي 1398-1387 اﺳﺖ. در راﺳﺘﺎي دﺳﺘﯿﺎﺑﯽ ﺑﻪ اﻫﺪاف ﭘﮋوﻫﺶ، اﺑﺘﺪا ﺑﺎ ﻣﻄﺎﻟﻌﻪ ﭘﮋوﻫﺶﻫﺎي ﭘﯿﺸﯿﻦ در ﺣﻮزه درﻣﺎﻧﺪﮔﯽ ﻣﺎﻟﯽ 12 ﻧﺴﺒﺖ ﻣﺎﻟﯽ اﺛﺮﮔﺬار ﺑﺮ ورﺷﮑﺴﺘﮕﯽ ﻣﺎﻟﯽ اﻧﺘﺨﺎب ﺷﺪه اﺳﺖ. ﭘﺲ از ﻣﺤﺎﺳﺒﻪ ﻧﺴﺒﺖﻫﺎ از آزﻣﻮن ﻣﻘﺎﯾﺴﻪ ﻣﯿﺎﻧﮕﯿﻦ اﺳﺘﻔﺎده ﺷﺪه اﺳﺖ ﺗﺎ ﻧﺴﺒﺖﻫﺎﯾﯽ ﮐﻪ ﺗﻔﺎوت ﻣﻌﻨﺎداري ﻣﯿﺎن دو ﮔﺮوه ورﺷﮑﺴﺘﻪ و ﻏﯿﺮورﺷﮑﺴﺘﻪ ﻣﺎﻟﯽ دارﻧﺪ، ﺑﺮاي ﻣﺤﺎﺳﺒﻪ در ﻣﺪلﻫﺎي ﭘﯿﺶﺑﯿﻨﯽ در ﻧﻈﺮ ﮔﺮﻓﺘﻪ ﺷﻮﻧﺪ ﮐﻪ ﻧﺘﺎﯾﺞ ﻧﺸﺎن داد ﻫﺮ 12 ﻣﺘﻐﯿﺮ ﺑﺮاي اﺳﺘﻔﺎده در ﻣﺪلﻫﺎ ﻣﻨﺎﺳﺐ ﻫﺴﺘﻨﺪ. ﺳﭙﺲ، ﺑﻪ ﻣﻨﻈﻮر ﺑﺮرﺳﯽ ﺗﻮان ﻣﺪﯾﺮﯾﺖ ﺳﺮﻣﺎﯾﻪ در ﮔﺮدش در ﭘﯿﺶﺑﯿﻨﯽ ورﺷﮑﺴﺘﮕﯽ ﻣﺎﻟﯽ ﺷﺮﮐﺖﻫﺎ، ﺑﻪ ﻣﻘﺎﯾﺴﻪ ﻣﺪلﻫﺎي ﭘﮋوﻫﺶ ﺑﺎ ﺗﻮﺟﻪ و ﺑﺪون ﺗﻮﺟﻪ ﺑﻪ ﻣﺘﻐﯿﺮ ﻣﺪﯾﺮﯾﺖ ﺳﺮﻣﺎﯾﻪ در ﮔﺮدش ﺑﺮ ﻣﺒﻨﺎي ﭘﻨﺞ ﻣﺪل ﺷﺒﮑﻪ ﻋﺼﺒﯽ ﭘﺮﺳﭙﺘﺮون ﭼﻨﺪ ﻻﯾﻪ، ﻣﺎﺷﯿﻦ ﺑﺮدار ﭘﺸﺘﯿﺒﺎن، درﺧﺖ ﺗﺼﻤﯿﻢ، رﮔﺮﺳﯿﻮن ﻟﺠﺴﺘﯿﮏ و ﺗﺤﻠﯿﻞ ﻣﻤﯿﺰي ﭼﻨﺪﮔﺎﻧﻪ ﭘﺮداﺧﺘﻪ ﺷﺪه اﺳﺖ. ﻧﺘﺎﯾﺞ ﻣﻘﺎﯾﺴﻪ ﻣﺪلﻫﺎي ﭘﯿﺶﺑﯿﻨﯽ ورﺷﮑﺴﺘﮕﯽ ﻧﺸﺎن داد ﻣﺪل ﺷﺒﮑﻪ ﻋﺼﺒﯽ ﭘﺮﺳﭙﺘﺮون ﭼﻨﺪ ﻻﯾﻪ، ﻧﺴﺒﺖ ﺑﻪ ﺳﺎﯾﺮ ﻣﺪلﻫﺎ داراي ﺑﯿﺸﺘﺮﯾﻦ ﻗﺪرت در ﭘﯿﺶﺑﯿﻨﯽ ﺷﺮﮐﺖﻫﺎ از ﻟﺤﺎظ ورﺷﮑﺴﺘﮕﯽ ﻣﺎﻟﯽ و ﺳﺎﻟﻢ ﺑﻮدن اﺳﺖ. ﻫﻤﭽﻨﯿﻦ، ﻧﺘﺎﯾﺞ ﻣﻘﺎﯾﺴﻪ ﻣﺪلﻫﺎ ﻧﺸﺎن داد ﺑﺎ ﺗﻮﺳﻌﻪ ﻣﺪل ﭘﮋوﻫﺶ، از ﻃﺮﯾﻖ وارد ﮐﺮدن ﻣﺘﻐﯿﺮ ﻣﺪﯾﺮﯾﺖ ﺳﺮﻣﺎﯾﻪ در ﮔﺮدش، ﺧﻄﺎي آﻣﻮزش ﻣﺪل ﺷﺒﮑﻪ ﻋﺼﺒﯽ ﭘﺮﺳﭙﺘﺮون ﭼﻨﺪ ﻻﯾﻪ ﺑـﻪ ﻣﻘـﺪار 0/036 ﮐـﺎﻫﺶ و ﺑﺮ دﻗﺖ ﻣﺪل ﺗﺎ 75 درﺻﺪ اﻓﺰوده ﻣﯽﺷﻮد.

The main purpose of this study is to model and determine the ability of working capital management in predicting financial bankruptcy of companies using artificial intelligence algorithms. The statistical population of the study consists of 120 companies listed on the Tehran Stock Exchange during the years 2008-2019. In order to achieve the objectives of the research, first by studying previous research in the field of financial distress, 12 financial ratios affecting financial bankruptcy have been selected. After calculating the ratios, the mean comparison test was used to consider the ratios that have a significant difference between the two bankrupt and non-bankrupt financial groups for calculation in the forecasting models, which showed that all 12 variables are suitable for use in the models. Then, in order to evaluate the ability of working capital management in predicting companies' financial bankruptcy, to compare research models with and without working capital management variable based on five models of multilayer perceptron neural network, support vector machine, decision tree, logistic regression and multiple audit analysis is performed. The results of comparing bankruptcy prediction models showed that the multilayer perceptron neural network model has the highest power in predicting companies in terms of financial bankruptcy and soundness compared to other models. The results of comparing the models showed that with the development of the research model, by entering the working capital management variable, the training error of the multilayer perceptron neural network model is reduced to 0.036 and the accuracy of the model is increased to 75%.