Dynamic forecasts of bankruptcy with Recurrent Neural Network model (RNN(Recurrent Neural Network)을 이용한 기업부도예측모형에서 회계정보의 동적 변화 연구)
-
- Journal of Intelligence and Information Systems
- /
- v.23 no.3
- /
- pp.139-153
- /
- 2017
Corporate bankruptcy can cause great losses not only to stakeholders but also to many related sectors in society. Through the economic crises, bankruptcy have increased and bankruptcy prediction models have become more and more important. Therefore, corporate bankruptcy has been regarded as one of the major topics of research in business management. Also, many studies in the industry are in progress and important. Previous studies attempted to utilize various methodologies to improve the bankruptcy prediction accuracy and to resolve the overfitting problem, such as Multivariate Discriminant Analysis (MDA), Generalized Linear Model (GLM). These methods are based on statistics. Recently, researchers have used machine learning methodologies such as Support Vector Machine (SVM), Artificial Neural Network (ANN). Furthermore, fuzzy theory and genetic algorithms were used. Because of this change, many of bankruptcy models are developed. Also, performance has been improved. In general, the company's financial and accounting information will change over time. Likewise, the market situation also changes, so there are many difficulties in predicting bankruptcy only with information at a certain point in time. However, even though traditional research has problems that don't take into account the time effect, dynamic model has not been studied much. When we ignore the time effect, we get the biased results. So the static model may not be suitable for predicting bankruptcy. Thus, using the dynamic model, there is a possibility that bankruptcy prediction model is improved. In this paper, we propose RNN (Recurrent Neural Network) which is one of the deep learning methodologies. The RNN learns time series data and the performance is known to be good. Prior to experiment, we selected non-financial firms listed on the KOSPI, KOSDAQ and KONEX markets from 2010 to 2016 for the estimation of the bankruptcy prediction model and the comparison of forecasting performance. In order to prevent a mistake of predicting bankruptcy by using the financial information already reflected in the deterioration of the financial condition of the company, the financial information was collected with a lag of two years, and the default period was defined from January to December of the year. Then we defined the bankruptcy. The bankruptcy we defined is the abolition of the listing due to sluggish earnings. We confirmed abolition of the list at KIND that is corporate stock information website. Then we selected variables at previous papers. The first set of variables are Z-score variables. These variables have become traditional variables in predicting bankruptcy. The second set of variables are dynamic variable set. Finally we selected 240 normal companies and 226 bankrupt companies at the first variable set. Likewise, we selected 229 normal companies and 226 bankrupt companies at the second variable set. We created a model that reflects dynamic changes in time-series financial data and by comparing the suggested model with the analysis of existing bankruptcy predictive models, we found that the suggested model could help to improve the accuracy of bankruptcy predictions. We used financial data in KIS Value (Financial database) and selected Multivariate Discriminant Analysis (MDA), Generalized Linear Model called logistic regression (GLM), Support Vector Machine (SVM), Artificial Neural Network (ANN) model as benchmark. The result of the experiment proved that RNN's performance was better than comparative model. The accuracy of RNN was high in both sets of variables and the Area Under the Curve (AUC) value was also high. Also when we saw the hit-ratio table, the ratio of RNNs that predicted a poor company to be bankrupt was higher than that of other comparative models. However the limitation of this paper is that an overfitting problem occurs during RNN learning. But we expect to be able to solve the overfitting problem by selecting more learning data and appropriate variables. From these result, it is expected that this research will contribute to the development of a bankruptcy prediction by proposing a new dynamic model.
There were strong criticisms against the joint development method: the redevelopment corporation and developers would achieve the whole development profit. The existing tenants who lost their housing in the site argued their right to reside in the site after the development was completed. There was also strong political pressure that the Roh Tae-woo governing administration should resolve the social inequality caused by the situation. In such circumstances, it was introduced that a certain proportion of public rental housing should be built in the redevelopment site; then the government took over the dwellings at a price of construction and allocated them to the existing tenants. The aims of this paper are to understand the rationale behind the inclusion of the public rental housing in the redevelopment sites; and to investigate to what extent the legislation was implemented appropriately. Although the legislation was introduced in Seoul from August 1989, it was not until May 2005 when it was implemented nationwide. At the beginning, there was an ambiguous rule that the number of public housing to be included should be limited to the number of households who would want to remain in the redeveloped site. In 2005 the Seoul metropolitan authority introduced a mandatory proportion; 17% of the total housing delivered in the site should be public rental homes. Since then the proportion. The proportion has been fluctuated by the political agenda of each ruling party: the conservative tended to reduce the proportion, whilst the opposition parties increased the proportion. Currently the proportion is 20% of the total stock to be built. Initially the size of the public housing was exceptionally small- less than 40 m2 but it has increased up to 60 m2 since 2010. The rental price was reasonably lower than market rent. The competition toward redevelopment rental housing that are vacant due to move or death of tenants was very high; it was given to one household out of nine eligible households in 2020.
Rapid growth of internet technology and social media is progressing. Data mining technology has evolved to enable unstructured document representations in a variety of applications. Sentiment analysis is an important technology that can distinguish poor or high-quality content through text data of products, and it has proliferated during text mining. Sentiment analysis mainly analyzes people's opinions in text data by assigning predefined data categories as positive and negative. This has been studied in various directions in terms of accuracy from simple rule-based to dictionary-based approaches using predefined labels. In fact, sentiment analysis is one of the most active researches in natural language processing and is widely studied in text mining. When real online reviews aren't available for others, it's not only easy to openly collect information, but it also affects your business. In marketing, real-world information from customers is gathered on websites, not surveys. Depending on whether the website's posts are positive or negative, the customer response is reflected in the sales and tries to identify the information. However, many reviews on a website are not always good, and difficult to identify. The earlier studies in this research area used the reviews data of the Amazon.com shopping mal, but the research data used in the recent studies uses the data for stock market trends, blogs, news articles, weather forecasts, IMDB, and facebook etc. However, the lack of accuracy is recognized because sentiment calculations are changed according to the subject, paragraph, sentiment lexicon direction, and sentence strength. This study aims to classify the polarity analysis of sentiment analysis into positive and negative categories and increase the prediction accuracy of the polarity analysis using the pretrained IMDB review data set. First, the text classification algorithm related to sentiment analysis adopts the popular machine learning algorithms such as NB (naive bayes), SVM (support vector machines), XGboost, RF (random forests), and Gradient Boost as comparative models. Second, deep learning has demonstrated discriminative features that can extract complex features of data. Representative algorithms are CNN (convolution neural networks), RNN (recurrent neural networks), LSTM (long-short term memory). CNN can be used similarly to BoW when processing a sentence in vector format, but does not consider sequential data attributes. RNN can handle well in order because it takes into account the time information of the data, but there is a long-term dependency on memory. To solve the problem of long-term dependence, LSTM is used. For the comparison, CNN and LSTM were chosen as simple deep learning models. In addition to classical machine learning algorithms, CNN, LSTM, and the integrated models were analyzed. Although there are many parameters for the algorithms, we examined the relationship between numerical value and precision to find the optimal combination. And, we tried to figure out how the models work well for sentiment analysis and how these models work. This study proposes integrated CNN and LSTM algorithms to extract the positive and negative features of text analysis. The reasons for mixing these two algorithms are as follows. CNN can extract features for the classification automatically by applying convolution layer and massively parallel processing. LSTM is not capable of highly parallel processing. Like faucets, the LSTM has input, output, and forget gates that can be moved and controlled at a desired time. These gates have the advantage of placing memory blocks on hidden nodes. The memory block of the LSTM may not store all the data, but it can solve the CNN's long-term dependency problem. Furthermore, when LSTM is used in CNN's pooling layer, it has an end-to-end structure, so that spatial and temporal features can be designed simultaneously. In combination with CNN-LSTM, 90.33% accuracy was measured. This is slower than CNN, but faster than LSTM. The presented model was more accurate than other models. In addition, each word embedding layer can be improved when training the kernel step by step. CNN-LSTM can improve the weakness of each model, and there is an advantage of improving the learning by layer using the end-to-end structure of LSTM. Based on these reasons, this study tries to enhance the classification accuracy of movie reviews using the integrated CNN-LSTM model.
1. Objective of the Study The objective of the study was to make fundamental suggestions for drawing a forest insurance system applicable in Korea by investigating forest insurance systems undertaken in foreign countries, analyzing the forest hazards occurred in entire forests of Korea in the past, and hearing the opinions of people engaged in forestry. 2. Methods of the Study First, reference studies on insurance at large as well as on forest insurance were intensively made to draw the characteristics of forest insurance practiced in main forestry countries, Second, the investigations of forest hazards in Korea for the past ten years were made with the help of the Office of Forestry. Third, the questionnaires concerning forest insurance were prepared and delivered at random to 533 personnel who are working at different administrative offices of forestry, forest stations, forest cooperatives, colleges and universities, research institutes, and fire insurance companies. Fourth, fifty three representative forest owners in the area of three forest types (coniferous, hardwood, and mixed forest), a representative region in Kyonggi Province out of fourteen collective forest development programs in Korea, were directly interviewed with the writer. 3. Results of the Study The rate of response to the questionnaire was 74.40% as shown in the table 3, and the results of the questionaire were as follows: (% in the parenthes shows the rates of response; shortages in amount to 100% were due to the facts of excluding the rates of response of minor respondents). 1) Necessity of forest insurance The respondents expressed their opinions that forest insurance must be undertaken to assure forest financing (5.65%); for receiving the reimbursement of replanting costs in case of damages done (35.87%); and to protect silvicultural investments (46.74%). 2) Law of forest insurance Few respondents showed their views in favor of applying the general insurance regulations to forest insurance practice (9.35%), but the majority of respondents were in favor of passing a special forest insurance law in the light of forest characteristics (88.26%). 3) Sorts of institutes to undertake forest insurance A few respondents believed that insurance companies at large could take care of forest insurance (17.42%); forest owner's mutual associations would manage the forest insurance more effectively (23.53%); but the more than half of the respondents were in favor of establishing public or national forest insurance institutes (56.18%). 4) Kinds of risks to be undertaken in forest insurance It would be desirable that the risks to be undertaken in forest insurance be limited: To forest fire hazards only (23.38%); to forest fire hazards plus damages made by weather (14.32%); to forest fire hazards, weather damages, and insect damages (60.68%). 5) Objectives to be insured It was responded that the objectives to be included in forest insurance should be limited: (1) To artificial coniferous forest only (13.47%); (2) to both coniferous and broad-leaved artificial forests (23.74%); (3) but the more than half of the respondents showed their desire that all the forests regardless of species and the methods of establishment should be insured (61.64%). 6) Range of risks in age of trees to be included in forest insurance The opinions of the respondents showed that it might be enough to insure the trees less than ten years of age (15.23%); but it would be more desirous of taking up forest trees under twenty years of age (32.95%); nevertheless, a large number of respondents were in favor of underwriting all the forest trees less than fourty years of age (46.37%). 7) Term of a forest insurance contract Quite a few respondents favored a contract made on one year basis (31.74%), but the more than half of the respondents favored the contract made on five year bases (58.68%). 8) Limitation in a forest insurance contract The respondents indicated that it would be desirable in a forest insurance contract to exclude forests less than five hectars (20.78%), but more than half of the respondents expressed their opinions that forests above a minimum volume or number of trees per unit area should be included in a forest insurance contract regardless of the area of forest lands (63.77%). 9) Methods of contract Some responded that it would be good to let the forest owners choose their forests in making a forest insurance contract (32.13%); others inclined to think that it would be desirable to include all the forests that owners hold whenerver they decide to make a forest insurance contract (33.48%); the rest responded in favor of forcing the owners to buy insurance policy if they own the forests that were established with subsidy or own highly vauable growing stock (31.92%) 10) Rate of premium The responses were divided into three categories: (1) The rate of primium is to be decided according to the regional degree of risks(27.72%); (2) to be decided by taking consideration both regional degree of risks and insurable values(31.59%); (3) and to be decided according to the rate of risks for the entire country and the insurable values (39.55%). 11) Payment of Premium Although a few respondents wished to make a payment of premium at once for a short term forest insurance contract, and an annual payment for a long term contract (13.80%); the majority of the respondents wished to pay the premium annually regardless of the term of contract, by employing a high rate of premium on a short term contract, but a low rate on a long term contract (83.71%). 12) Institutes in charge of forest insurance business A few respondents showed their desire that forest insurance be taken care of at the government forest administrative offices (18.75%); others at insurance companies (35.76%); but the rest, the largest number of the respondents, favored forest associations in the county. They also wanted to pay a certain rate of premium to the forest associations that issue the insurance (44.22%). 13) Limitation on indemnity for damages done In limitation on indemnity for damages done, the respondents showed a quite different views. Some desired compesation to cover replanting costs when young stands suffered damages and to be paid at the rate of eighty percent to the losses received when matured timber stands suffered damages(29.70%); others desired to receive compensation of the actual total loss valued at present market prices (31.07%); but the rest responded in favor of compensation at the present value figured out by applying a certain rate of prolongation factors to the establishment costs(36.99%). 14) Raising of funds for forest insurance A few respondents hoped to raise the fund for forest insurance by setting aside certain amount of money from the indemnity paid (15.65%); others wished to raise the fund by levying new forest land taxes(33.79%); but the rest expressed their hope to raise the fund by reserving certain amount of money from the surplus money that was saved due to the non-risks (44.81%). 15) Causes of fires The main causes of forest fires 6gured out by the respondents experience turned out to be (1) an accidental fire, (2) cigarettes, (3) shifting cultivation. The reponses were coincided with the forest fire analysis made by the Office of Forestry. 16) Fire prevention The respondents suggested that the most important and practical three kinds of forest fire prevention measures would be (1) providing a fire-break, (2) keeping passers-by out during the drought seasons, (3) enlightenment through mass communication systems. 4. Suggestions The writer wishes to present some suggestions that seemed helpful in drawing up a forest insurance system by reviewing the findings in the questionaire analysis and the results of investigations on forest insurance undertaken in foreign countries. 1) A forest insurance system designed to compensate the loss figured out on the basis of replanting cost when young forest stands suffered damages, and to strengthen credit rating by relieving of risks of damages, must be put in practice as soon as possible with the enactment of a specifically drawn forest insurance law. And the committee of forest insurance should be organized to make a full study of forest insurance system. 2) Two kinds of forest insurance organizations furnishing forest insurance, publicly-owned insurance organizations and privately-owned, are desirable in order to handle forest risks properly. The privately-owned forest insurance organizations should take up forest fire insurance only, and the publicly-owned ought to write insurance for forest fires and insect damages. 3) The privately-owned organizations furnishing forest insurance are desired to take up all the forest stands older than twenty years; whereas, the publicly-owned should sell forest insurance on artificially planted stands younger than twenty years with emphasis on compensating replanting costs of forest stands when they suffer damages. 4) Small forest stands, less than one hectare holding volume or stocked at smaller than standard per unit area are not to be included in a forest insurance writing, and the minimum term of insuring should not be longer than one year in the privately-owned forest insurance organizations although insuring period could be extended more than one year; whereas, consecutive five year term of insurance periods should be set as a mimimum period of insuring forest in the publicly-owned forest insurance organizations. 5) The forest owners should be free in selecting their forests in insuring; whereas, forest owners of the stands that were established with subsidy should be required to insure their forests at publicly-owned forest insurance organizations. 6) Annual insurance premiums for both publicly-owned and privately-owned forest insurance organizations ought to be figured out in proportion to the amount of insurance in accordance with the degree of risks which are grouped into three categories on the basis of the rate of risks throughout the country. 7) Annual premium should be paid at the beginning of forest insurance contract, but reduction must be made if the insuring periods extend longer than a minimum period of forest insurance set by the law. 8) The compensation for damages, the reimbursement, should be figured out on the basis of the ratio between the amount of insurance and insurable value. In the publicly-owned forest insurance system, the standard amount of insurance should be set on the basis of establishment costs in order to prevent over-compensation. 9) Forest insurance business is to be taken care of at the window of insurance com pnies when forest owners buy the privately-owned forest insurance, but the business of writing the publicly-owned forest insurance should be done through the forest cooperatives and certain portions of the premium be reimbursed to the forest cooperatives. 10) Forest insurance funds ought to be reserved by levying a property tax on forest lands. 11) In order to prevent forest damages, the forest owners should be required to report forest hazards immediately to the forest insurance organizations and the latter should bear the responsibility of taking preventive measures.