Research On Financial Distress Prediction Method Of Listed Companies With The Introduction Of Financial Status Analysis

The listed companies of China constitute the lifeblood of the national economy.The company’s financial status determines the company’s development trend and even the country’s economic trend.The study of financial distress prediction provides an analytical tool for discovering the company’s financial deterioration signal in advance.The core content of the study is to reveal the relationship between the company’s financial data and the risk of financial distress,and to determine the possibility of a company falling into a financial distress in the future.Therefore,analyzing the development law of the company’s financial status and forecasting the event of financial distress have become the core issues of academic and industrial.At present,a myraid of research have been achieved in the field of financial distress prediction,but most of the research is focused on the design and construction of the financial distress prediction model itself.The financial status information contained in the data set and the role of this kind of information in the construction of the financial distress prediction model have not been fully investigated.Therefore,this study starts with the analysis of financial status information,and studies the role and modeling function of financial status information in the face of small sample financial datasets,large sample financial datasets and financial longitudinal data stream datasets.In the specific research process,the role of financial status space optimization,multi-financial status partitioning and financial path information based on financial status sequence analysis are investigated.This paper mainly includes three aspects: the construction of financial distress prediction combination model based on financial status space optimization,the construction of two-layer selective ensemble model based on multi-financial status partition,and the construction of financial distress prediction model based on financial status sequence analysis.This study is made up of six chapters.The first chapter is the introduction;the second chapter is the theoretical basis of the firm financial distress prediction;the third chapter is the selection of experimental data and financial indicator system;the fourth chapter is combination method of financial distrss prediction based on financial status space optimization,and the fifth chapter is two-layer selective ensemble of financial distrss prediction based on multi-financial status partition analysis.The sixth chapter is financial distress prediction based on financial status sequence analysis.The main contents include three aspects:(1)First,a financial distress prediction combination model based on financial status space optimization is constructed.When facing the small sample dataset,the sample data distribution in the financial data set is relatively simple.In order to preserve the effective classification information in the data set,the training model is immune to the influence of noise data and outlier samples.Firstly,this paper analyzes the necessity of sample selection in the establishment of financial distress prediction model.Then,according to the actual situation of two types of sample division of listed companies in China,the method based on fuzzy clustering is adopted to select the financial sample set,and the samples which do not conform to the divided financial status are removed.Because probabilistic neural network has the characteristics of fast convergence and simple training process,it is used as the basic classifier of financial distress prediction.At the same time,because the volume of data is relatively small,particle swarm optimization is used to optimize the parameters of fuzzy clustering method and probabilistic neural network.Empirical results show that the model can effectively retain the classification information in the initial financial dataset,and the financial distress prediction combination model based on the optimization of financial status space is more suitable for the prediction of small sample financial data set.(2)Secondly,a two-layer selective ensemble financial distress prediction model based on multi-financial status partition is constructed.Because the large sample financial dataset is composed of a long time span of financial sample sets,so the data set contains a various types of classification information and status space.In order to distinguish the status contained in the financial dataset purposefully and ensure that each subset of the partitioned data contains specific financial status information for classification,this paper studies the relationship between the quantitative analysis method and the classifier ensemble method,and proposes a two-layer selective ensemble financial distress prediction method.In the process of modeling,in order to ensure the diversity between the base classifiers,this paper uses three kernel-based fuzzy c-means methods to group the financial dataset.Then,in order to overcome the negtive prediction effect caused by the integration of poor performance classifiers,the forward selection ensemble method is used to integrate each group of classifiers separately.Finally,in order to meet the requirements of the ensemble model for classifier performance,a twolayer ensemble method is adopted to implement the condition.The empirical results show that the multi-financial status partition can effectively distinguish the different classified information contained in the financial dataset with multiclassification space,and ensure the diversity between the base classifiers.The two-layer selective ensemble model achieves better prediction results.(3)Finally,a financial distress prediction model based on financial status sequence analysis is proposed.When considering the time factor,the dataset of financial longitudinal data stream provides the condition for analyzing the changing process of financial status.According to the separability between financial distress companies and financial health companies,the classification model can be established by using the financial path information of the two types of companies.In order to quantify the changing process of financial status,this study introduces the concept of "financial path".Because the fuzzy clustering method can classify the financial data sets according to the characteristics of financial indicators,this study uses the accounting year as a timestamp to classify the financial data sets of each year in the financial data stream into multi-financial status groups,and designs the financial status sequence experienced by the company in the past few years.Based on the basic idea that the financial paths experienced by financial distress companies and non-financial distress companies are separable,the classifier model is designed according to the Euclidean distance between paths after the financial paths are designed.Empirical results show that,on the one hand,the classifier model designed in this paper can visually analyze the changing process of financial state,ensure that the financial distress prediction is no longer a "black box" model,and provide a process to interpret the occurrence of financial distress.On the other hand,the historical financial datasets can be fully used by the mode.The financial distress prediction model based on financial status sequence analysis has a good performance in predicting financial distress.