A New Idea of Study on the Influence Factors of Companies' Debt Costs in the Big Data Era

Under the background of big data era today, once been widely used method – multiple linear regressions can not satisfy people’s need to handle big data any more because of its bad characteristics such as multicollinearity, instability, subjectivity in model chosen etc. Contrary to MLR, LASSO method has many good natures. it is stable and can handle multicollinearity and successfully select the best model and do estimation in the same time. LASSO method is an effective improvement of multiple linear regressions. It is a natural change and innovation to introduce LASSO method into the accounting field and use it to deal with the debt costs problems. It helps us join the statistic field and accounting field together step by step. What’s more, in order to proof the applicability of LASSO method in dealing with debt costs problems, we take 2301 companies’ data from Shanghai and Shenzhen A-share market in 2012 as samples, and chose 18 indexes to verify that the results of LASSO method is scientific, reasonable and accurate. In the end, we compare LASSO method with traditional multiple linear regressions and ridge regression, finding out that LASSO method can not only offer the most accurate prediction but also simplify the model. © 2014 The Authors. Published by Elsevier B.V. Selection and/or peer-review under responsibility of the organizers of ITQM 2014