به كارگيري شبكه‌هاي عصبي در پيش‌بيني سودآوري شركتها (شركتهاي عضو بورس اوراق بهادار تهران)

Application of Neural Networks In Corporate’s Pro?tability Prediction

اين پژوهش در مورد دستيابي به پيش‌بيني سودآوري آينده (شاخصي از عملكرد آينده) شركت‌هاي پذيرفته شده در بورس اوراق بهادار تهران با كاربست شبكه‌هاي عصبي مصنوعي است. نمونه‌هاي مورد استفاده ما در اين پژوهش 90 شركت از مجموعه مورد اشاره فوق در بين سال‌هاي 1380 تا 1386 (720 شركت - سال) مي‌باشد و با توجه به چارچوب مقاله اين تعداد به 630 شركت - سال كاهش يافت. اين شركتها به دو مجموعه نمونه آموزشي شامل 540 شركت – سال ابتدايي (شركت – سالهاي 80 تا 85) و براي ارزيابي اعتبار الگوي آموزش ديده، نمونه آزمايشي متشكل از 90 شركت – سال انتهايي (شركت – سال 86)، تفكيك شدند.

This study aims at profitability prediction of listed companies in Tehran Stocks Exchange (TSE), using Artificial Neural Network. The respected sample consists of 90 firms from 2002 to 2009 (720 firm/years). Attention to the framework of study reduced the number of 720 firm/years to 630 firm/years. These firms separated in two groups of learning sample (540 firm/years) and test sample (90 firm/years) to test generalization of the technique. To develop profitability prediction, first, we needed to determine predictor variables. Profitability prediction literature was reviewed and a complete list of financial ratios for successful prediction in the past studies was prepared. Then, we reduced the list from a theoretical point of view, and we used SDA technique to select final financial ratios. Finally, we took 9 financial ratios to develop profitability prediction. Using Artificial Neural Network (ANN) and applying 9 selected financial ratios, achieved 99% accuracy rate in the learning sample and 86% accuracy rate in the test sample for correct classification of the firms into profitability and nonprofitability groups one year before the real state.