Optimization of Multiclass Support Vector Machine using Genetic Algorithm: Application to the Prediction of Corporate Credit Rating (유전자 알고리즘을 이용한 다분류 SVM의 최적화: 기업신용등급 예측에의 응용)
-
- Information Systems Review
- /
- v.16 no.3
- /
- pp.161-177
- /
- 2014
Corporate credit rating assessment consists of complicated processes in which various factors describing a company are taken into consideration. Such assessment is known to be very expensive since domain experts should be employed to assess the ratings. As a result, the data-driven corporate credit rating prediction using statistical and artificial intelligence (AI) techniques has received considerable attention from researchers and practitioners. In particular, statistical methods such as multiple discriminant analysis (MDA) and multinomial logistic regression analysis (MLOGIT), and AI methods including case-based reasoning (CBR), artificial neural network (ANN), and multiclass support vector machine (MSVM) have been applied to corporate credit rating.2) Among them, MSVM has recently become popular because of its robustness and high prediction accuracy. In this study, we propose a novel optimized MSVM model, and appy it to corporate credit rating prediction in order to enhance the accuracy. Our model, named 'GAMSVM (Genetic Algorithm-optimized Multiclass Support Vector Machine),' is designed to simultaneously optimize the kernel parameters and the feature subset selection. Prior studies like Lorena and de Carvalho (2008), and Chatterjee (2013) show that proper kernel parameters may improve the performance of MSVMs. Also, the results from the studies such as Shieh and Yang (2008) and Chatterjee (2013) imply that appropriate feature selection may lead to higher prediction accuracy. Based on these prior studies, we propose to apply GAMSVM to corporate credit rating prediction. As a tool for optimizing the kernel parameters and the feature subset selection, we suggest genetic algorithm (GA). GA is known as an efficient and effective search method that attempts to simulate the biological evolution phenomenon. By applying genetic operations such as selection, crossover, and mutation, it is designed to gradually improve the search results. Especially, mutation operator prevents GA from falling into the local optima, thus we can find the globally optimal or near-optimal solution using it. GA has popularly been applied to search optimal parameters or feature subset selections of AI techniques including MSVM. With these reasons, we also adopt GA as an optimization tool. To empirically validate the usefulness of GAMSVM, we applied it to a real-world case of credit rating in Korea. Our application is in bond rating, which is the most frequently studied area of credit rating for specific debt issues or other financial obligations. The experimental dataset was collected from a large credit rating company in South Korea. It contained 39 financial ratios of 1,295 companies in the manufacturing industry, and their credit ratings. Using various statistical methods including the one-way ANOVA and the stepwise MDA, we selected 14 financial ratios as the candidate independent variables. The dependent variable, i.e. credit rating, was labeled as four classes: 1(A1); 2(A2); 3(A3); 4(B and C). 80 percent of total data for each class was used for training, and remaining 20 percent was used for validation. And, to overcome small sample size, we applied five-fold cross validation to our dataset. In order to examine the competitiveness of the proposed model, we also experimented several comparative models including MDA, MLOGIT, CBR, ANN and MSVM. In case of MSVM, we adopted One-Against-One (OAO) and DAGSVM (Directed Acyclic Graph SVM) approaches because they are known to be the most accurate approaches among various MSVM approaches. GAMSVM was implemented using LIBSVM-an open-source software, and Evolver 5.5-a commercial software enables GA. Other comparative models were experimented using various statistical and AI packages such as SPSS for Windows, Neuroshell, and Microsoft Excel VBA (Visual Basic for Applications). Experimental results showed that the proposed model-GAMSVM-outperformed all the competitive models. In addition, the model was found to use less independent variables, but to show higher accuracy. In our experiments, five variables such as X7 (total debt), X9 (sales per employee), X13 (years after founded), X15 (accumulated earning to total asset), and X39 (the index related to the cash flows from operating activity) were found to be the most important factors in predicting the corporate credit ratings. However, the values of the finally selected kernel parameters were found to be almost same among the data subsets. To examine whether the predictive performance of GAMSVM was significantly greater than those of other models, we used the McNemar test. As a result, we found that GAMSVM was better than MDA, MLOGIT, CBR, and ANN at the 1% significance level, and better than OAO and DAGSVM at the 5% significance level.
In franchise business, exclusive sales territory (sometimes EST in table) protection is a very important issue from an economic, social and political point of view. It affects the growth and survival of both franchisor and franchisee and often raises issues of social and political conflicts. When franchisee is not familiar with related laws and regulations, franchisor has high chance to utilize it. Exclusive sales territory protection by the manufacturer and distributors (wholesalers or retailers) means sales area restriction by which only certain distributors have right to sell products or services. The distributor, who has been granted exclusive sales territories, can protect its own territory, whereas he may be prohibited from entering in other regions. Even though exclusive sales territory is a quite critical problem in franchise business, there is not much rigorous research about the reason, results, evaluation, and future direction based on empirical data. This paper tries to address this problem not only from logical and nomological validity, but from empirical validation. While we purse an empirical analysis, we take into account the difficulties of real data collection and statistical analysis techniques. We use a set of disclosure document data collected by Korea Fair Trade Commission, instead of conventional survey method which is usually criticized for its measurement error. Existing theories about exclusive sales territory can be summarized into two groups as shown in the table below. The first one is about the effectiveness of exclusive sales territory from both franchisor and franchisee point of view. In fact, output of exclusive sales territory can be positive for franchisors but negative for franchisees. Also, it can be positive in terms of sales but negative in terms of profit. Therefore, variables and viewpoints should be set properly. The other one is about the motive or reason why exclusive sales territory is protected. The reasons can be classified into four groups - industry characteristics, franchise systems characteristics, capability to maintain exclusive sales territory, and strategic decision. Within four groups of reasons, there are more specific variables and theories as below. Based on these theories, we develop nine hypotheses which are briefly shown in the last table below with the results. In order to validate the hypothesis, data is collected from government (FTC) homepage which is open source. The sample consists of 1,896 franchisors and it contains about three year operation data, from 2006 to 2008. Within the samples, 627 have exclusive sales territory protection policy and the one with exclusive sales territory policy is not evenly distributed over 19 representative industries. Additional data are also collected from another government agency homepage, like Statistics Korea. Also, we combine data from various secondary sources to create meaningful variables as shown in the table below. All variables are dichotomized by mean or median split if they are not inherently dichotomized by its definition, since each hypothesis is composed by multiple variables and there is no solid statistical technique to incorporate all these conditions to test the hypotheses. This paper uses a simple chi-square test because hypotheses and theories are built upon quite specific conditions such as industry type, economic condition, company history and various strategic purposes. It is almost impossible to find all those samples to satisfy them and it can't be manipulated in experimental settings. However, more advanced statistical techniques are very good on clean data without exogenous variables, but not good with real complex data. The chi-square test is applied in a way that samples are grouped into four with two criteria, whether they use exclusive sales territory protection or not, and whether they satisfy conditions of each hypothesis. So the proportion of sample franchisors which satisfy conditions and protect exclusive sales territory, does significantly exceed the proportion of samples that satisfy condition and do not protect. In fact, chi-square test is equivalent with the Poisson regression which allows more flexible application. As results, only three hypotheses are accepted. When attitude toward the risk is high so loyalty fee is determined according to sales performance, EST protection makes poor results as expected. And when franchisor protects EST in order to recruit franchisee easily, EST protection makes better results. Also, when EST protection is to improve the efficiency of franchise system as a whole, it shows better performances. High efficiency is achieved as EST prohibits the free riding of franchisee who exploits other's marketing efforts, and it encourages proper investments and distributes franchisee into multiple regions evenly. Other hypotheses are not supported in the results of significance testing. Exclusive sales territory should be protected from proper motives and administered for mutual benefits. Legal restrictions driven by the government agency like FTC could be misused and cause mis-understandings. So there need more careful monitoring on real practices and more rigorous studies by both academicians and practitioners.
This study analyses the difference of contents and tones of arguments among three Korean major newspapers, the Kyunghyang Shinmoon, the HanKyoreh, and the Dong-A Ilbo. It is commonly accepted that newspapers in Korea explicitly deliver their own tone of arguments when they talk about some sensitive issues and topics. It could be controversial if readers of newspapers read the news without being aware of the type of tones of arguments because the contents and the tones of arguments can affect readers easily. Thus it is very desirable to have a new tool that can inform the readers of what tone of argument a newspaper has. This study presents the results of clustering and classification techniques as part of text mining analysis. We focus on six main subjects such as Culture, Politics, International, Editorial-opinion, Eco-business and National issues in newspapers, and attempt to identify differences and similarities among the newspapers. The basic unit of text mining analysis is a paragraph of news articles. This study uses a keyword-network analysis tool and visualizes relationships among keywords to make it easier to see the differences. Newspaper articles were gathered from KINDS, the Korean integrated news database system. KINDS preserves news articles of the Kyunghyang Shinmun, the HanKyoreh and the Dong-A Ilbo and these are open to the public. This study used these three Korean major newspapers from KINDS. About 3,030 articles from 2008 to 2012 were used. International, national issues and politics sections were gathered with some specific issues. The International section was collected with the keyword of 'Nuclear weapon of North Korea.' The National issues section was collected with the keyword of '4-major-river.' The Politics section was collected with the keyword of 'Tonghap-Jinbo Dang.' All of the articles from April 2012 to May 2012 of Eco-business, Culture and Editorial-opinion sections were also collected. All of the collected data were handled and edited into paragraphs. We got rid of stop-words using the Lucene Korean Module. We calculated keyword co-occurrence counts from the paired co-occurrence list of keywords in a paragraph. We made a co-occurrence matrix from the list. Once the co-occurrence matrix was built, we used the Cosine coefficient matrix as input for PFNet(Pathfinder Network). In order to analyze these three newspapers and find out the significant keywords in each paper, we analyzed the list of 10 highest frequency keywords and keyword-networks of 20 highest ranking frequency keywords to closely examine the relationships and show the detailed network map among keywords. We used NodeXL software to visualize the PFNet. After drawing all the networks, we compared the results with the classification results. Classification was firstly handled to identify how the tone of argument of a newspaper is different from others. Then, to analyze tones of arguments, all the paragraphs were divided into two types of tones, Positive tone and Negative tone. To identify and classify all of the tones of paragraphs and articles we had collected, supervised learning technique was used. The Na
The government recently announced various policies for developing big-data and artificial intelligence fields to provide a great opportunity to the public with respect to disclosure of high-quality data within public institutions. KSURE(Korea Trade Insurance Corporation) is a major public institution for financial policy in Korea, and thus the company is strongly committed to backing export companies with various systems. Nevertheless, there are still fewer cases of realized business model based on big-data analyses. In this situation, this paper aims to develop a new business model which can be applied to an ex-ante prediction for the likelihood of the insurance accident of credit guarantee. We utilize internal data from KSURE which supports export companies in Korea and apply machine learning models. Then, we conduct performance comparison among the predictive models including Logistic Regression, Random Forest, XGBoost, LightGBM, and DNN(Deep Neural Network). For decades, many researchers have tried to find better models which can help to predict bankruptcy since the ex-ante prediction is crucial for corporate managers, investors, creditors, and other stakeholders. The development of the prediction for financial distress or bankruptcy was originated from Smith(1930), Fitzpatrick(1932), or Merwin(1942). One of the most famous models is the Altman's Z-score model(Altman, 1968) which was based on the multiple discriminant analysis. This model is widely used in both research and practice by this time. The author suggests the score model that utilizes five key financial ratios to predict the probability of bankruptcy in the next two years. Ohlson(1980) introduces logit model to complement some limitations of previous models. Furthermore, Elmer and Borowski(1988) develop and examine a rule-based, automated system which conducts the financial analysis of savings and loans. Since the 1980s, researchers in Korea have started to examine analyses on the prediction of financial distress or bankruptcy. Kim(1987) analyzes financial ratios and develops the prediction model. Also, Han et al.(1995, 1996, 1997, 2003, 2005, 2006) construct the prediction model using various techniques including artificial neural network. Yang(1996) introduces multiple discriminant analysis and logit model. Besides, Kim and Kim(2001) utilize artificial neural network techniques for ex-ante prediction of insolvent enterprises. After that, many scholars have been trying to predict financial distress or bankruptcy more precisely based on diverse models such as Random Forest or SVM. One major distinction of our research from the previous research is that we focus on examining the predicted probability of default for each sample case, not only on investigating the classification accuracy of each model for the entire sample. Most predictive models in this paper show that the level of the accuracy of classification is about 70% based on the entire sample. To be specific, LightGBM model shows the highest accuracy of 71.1% and Logit model indicates the lowest accuracy of 69%. However, we confirm that there are open to multiple interpretations. In the context of the business, we have to put more emphasis on efforts to minimize type 2 error which causes more harmful operating losses for the guaranty company. Thus, we also compare the classification accuracy by splitting predicted probability of the default into ten equal intervals. When we examine the classification accuracy for each interval, Logit model has the highest accuracy of 100% for 0~10% of the predicted probability of the default, however, Logit model has a relatively lower accuracy of 61.5% for 90~100% of the predicted probability of the default. On the other hand, Random Forest, XGBoost, LightGBM, and DNN indicate more desirable results since they indicate a higher level of accuracy for both 0~10% and 90~100% of the predicted probability of the default but have a lower level of accuracy around 50% of the predicted probability of the default. When it comes to the distribution of samples for each predicted probability of the default, both LightGBM and XGBoost models have a relatively large number of samples for both 0~10% and 90~100% of the predicted probability of the default. Although Random Forest model has an advantage with regard to the perspective of classification accuracy with small number of cases, LightGBM or XGBoost could become a more desirable model since they classify large number of cases into the two extreme intervals of the predicted probability of the default, even allowing for their relatively low classification accuracy. Considering the importance of type 2 error and total prediction accuracy, XGBoost and DNN show superior performance. Next, Random Forest and LightGBM show good results, but logistic regression shows the worst performance. However, each predictive model has a comparative advantage in terms of various evaluation standards. For instance, Random Forest model shows almost 100% accuracy for samples which are expected to have a high level of the probability of default. Collectively, we can construct more comprehensive ensemble models which contain multiple classification machine learning models and conduct majority voting for maximizing its overall performance.
With the MICE(Meeting, Incentive travel, Convention, Exhibition) industry coming into the spotlight, there has been a growing interest in the domestic exhibition industry. Accordingly, in Korea, various studies of the industry are being conducted to enhance exhibition performance as in the United States or Europe. Some studies are focusing particularly on analyzing visiting patterns of exhibition visitors using intelligent information technology in consideration of the variations in effects of watching exhibitions according to the exhibitory environment or technique, thereby understanding visitors and, furthermore, drawing the correlations between exhibiting businesses and improving exhibition performance. However, previous studies related to booth recommendation systems only discussed the accuracy of recommendation in the aspect of a system rather than determining changes in visitors' behavior or perception by recommendation. A booth recommendation system enables visitors to visit unplanned exhibition booths by recommending visitors suitable ones based on information about visitors' visits. Meanwhile, some visitors may be satisfied with their unplanned visits, while others may consider the recommending process to be cumbersome or obstructive to their free observation. In the latter case, the exhibition is likely to produce worse results compared to when visitors are allowed to freely observe the exhibition. Thus, in order to apply a booth recommendation system to exhibition halls, the factors affecting the performance of the system should be generally examined, and the effects of the system on visitors' unplanned visiting behavior should be carefully studied. As such, this study aims to determine the factors that affect the performance of a booth recommendation system by reviewing theories and literature and to examine the effects of visitors' perceived performance of the system on their satisfaction of unplanned behavior and intention to reuse the system. Toward this end, the unplanned behavior theory was adopted as the theoretical framework. Unplanned behavior can be defined as "behavior that is done by consumers without any prearranged plan". Thus far, consumers' unplanned behavior has been studied in various fields. The field of marketing, in particular, has focused on unplanned purchasing among various types of unplanned behavior, which has been often confused with impulsive purchasing. Nevertheless, the two are different from each other; while impulsive purchasing means strong, continuous urges to purchase things, unplanned purchasing is behavior with purchasing decisions that are made inside a store, not before going into one. In other words, all impulsive purchases are unplanned, but not all unplanned purchases are impulsive. Then why do consumers engage in unplanned behavior? Regarding this question, many scholars have made many suggestions, but there has been a consensus that it is because consumers have enough flexibility to change their plans in the middle instead of developing plans thoroughly. In other words, if unplanned behavior costs much, it will be difficult for consumers to change their prearranged plans. In the case of the exhibition hall examined in this study, visitors learn the programs of the hall and plan which booth to visit in advance. This is because it is practically impossible for visitors to visit all of the various booths that an exhibition operates due to their limited time. Therefore, if the booth recommendation system proposed in this study recommends visitors booths that they may like, they can change their plans and visit the recommended booths. Such visiting behavior can be regarded similarly to consumers' visit to a store or tourists' unplanned behavior in a tourist spot and can be understand in the same context as the recent increase in tourism consumers' unplanned behavior influenced by information devices. Thus, the following research model was established. This research model uses visitors' perceived performance of a booth recommendation system as the parameter, and the factors affecting the performance include trust in the system, exhibition visitors' knowledge levels, expected personalization of the system, and the system's threat to freedom. In addition, the causal relation between visitors' satisfaction of their perceived performance of the system and unplanned behavior and their intention to reuse the system was determined. While doing so, trust in the booth recommendation system consisted of 2nd order factors such as competence, benevolence, and integrity, while the other factors consisted of 1st order factors. In order to verify this model, a booth recommendation system was developed to be tested in 2011 DMC Culture Open, and 101 visitors were empirically studied and analyzed. The results are as follows. First, visitors' trust was the most important factor in the booth recommendation system, and the visitors who used the system perceived its performance as a success based on their trust. Second, visitors' knowledge levels also had significant effects on the performance of the system, which indicates that the performance of a recommendation system requires an advance understanding. In other words, visitors with higher levels of understanding of the exhibition hall learned better the usefulness of the booth recommendation system. Third, expected personalization did not have significant effects, which is a different result from previous studies' results. This is presumably because the booth recommendation system used in this study did not provide enough personalized services. Fourth, the recommendation information provided by the booth recommendation system was not considered to threaten or restrict one's freedom, which means it is valuable in terms of usefulness. Lastly, high performance of the booth recommendation system led to visitors' high satisfaction levels of unplanned behavior and intention to reuse the system. To sum up, in order to analyze the effects of a booth recommendation system on visitors' unplanned visits to a booth, empirical data were examined based on the unplanned behavior theory and, accordingly, useful suggestions for the establishment and design of future booth recommendation systems were made. In the future, further examination should be conducted through elaborate survey questions and survey objects.
Digital Convergence means integration between industry, technology, and contents, and in marketing, it usually comes with creation of new types of product and service under the base of digital technology as digitalization progress in electro-communication industries including telecommunication, home appliance, and computer industries. One can see digital convergence not only in instruments such as PC, AV appliances, cellular phone, but also in contents, network, service that are required in production, modification, distribution, re-production of information. Convergence in contents started around 1990. Convergence in network and service begins as broadcasting and telecommunication integrates and DMB(digital multimedia broadcasting), born in May, 2005 is the symbolic icon in this trend. There are some positive and negative expectations about DMB. The reason why two opposite expectations exist is that DMB does not come out from customer's need but from technology development. Therefore, customers might have hard time to interpret the real meaning of DMB. Time is quite critical to a high tech product, like DMB because another product with same function from different technology can replace the existing product within short period of time. If DMB does not positioning well to customer's mind quickly, another products like Wibro, IPTV, or HSPDA could replace it before it even spreads out. Therefore, positioning strategy is critical for success of DMB product. To make correct positioning strategy, one needs to understand how consumer interprets DMB and how consumer's interpretation can be changed via communication strategy. In this study, we try to investigate how consumer perceives a new product, like DMB and how AD strategy change consumer's perception. More specifically, the paper segment consumers into sub-groups based on their DMB perceptions and compare their characteristics in order to understand how they perceive DMB. And, expose them different printed ADs that have messages guiding consumer think DMB in specific ways, either cellular phone or personal TV. Research Question 1: Segment consumers according to perceptions about DMB and compare characteristics of segmentations. Research Question 2: Compare perceptions about DMB after AD that induces categorization of DMB in direction for each segment. If one understand and predict a direction in which consumer perceive a new product, firm can select target customers easily. We segment consumers according to their perception and analyze characteristics in order to find some variables that can influence perceptions, like prior experience, usage, or habit. And then, marketing people can use this variables to identify target customers and predict their perceptions. If one knows how customer's perception is changed via AD message, communication strategy could be constructed properly. Specially, information from segmented customers helps to develop efficient AD strategy for segment who has prior perception. Research framework consists of two measurements and one treatment, O1 X O2. First observation is for collecting information about consumer's perception and their characteristics. Based on first observation, the paper segment consumers into two groups, one group perceives DMB similar to Cellular phone and the other group perceives DMB similar to TV. And compare characteristics of two segments in order to find reason why they perceive DMB differently. Next, we expose two kinds of AD to subjects. One AD describes DMB as Cellular phone and the other Ad describes DMB as personal TV. When two ADs are exposed to subjects, consumers don't know their prior perception of DMB, in other words, which subject belongs 'similar-to-Cellular phone' segment or 'similar-to-TV' segment? However, we analyze the AD's effect differently for each segment. In research design, final observation is for investigating AD effect. Perception before AD is compared with perception after AD. Comparisons are made for each segment and for each AD. For the segment who perceives DMB similar to TV, AD that describes DMB as cellular phone could change the prior perception. And AD that describes DMB as personal TV, could enforce the prior perception. For data collection, subjects are selected from undergraduate students because they have basic knowledge about most digital equipments and have open attitude about a new product and media. Total number of subjects is 240. In order to measure perception about DMB, we use indirect measurement, comparison with other similar digital products. To select similar digital products, we pre-survey students and then finally select PDA, Car-TV, Cellular Phone, MP3 player, TV, and PSP. Quasi experiment is done at several classes under instructor's allowance. After brief introduction, prior knowledge, awareness, and usage about DMB as well as other digital instruments is asked and their similarities and perceived characteristics are measured. And then, two kinds of manipulated color-printed AD are distributed and similarities and perceived characteristics for DMB are re-measured. Finally purchase intension, AD attitude, manipulation check, and demographic variables are asked. Subjects are given small gift for participation. Stimuli are color-printed advertising. Their actual size is A4 and made after several pre-test from AD professionals and students. As results, consumers are segmented into two subgroups based on their perceptions of DMB. Similarity measure between DMB and cellular phone and similarity measure between DMB and TV are used to classify consumers. If subject whose first measure is less than the second measure, she is classified into segment A and segment A is characterized as they perceive DMB like TV. Otherwise, they are classified as segment B, who perceives DMB like cellular phone. Discriminant analysis on these groups with their characteristics of usage and attitude shows that Segment A knows much about DMB and uses a lot of digital instrument. Segment B, who thinks DMB as cellular phone doesn't know well about DMB and not familiar with other digital instruments. So, consumers with higher knowledge perceive DMB similar to TV because launching DMB advertising lead consumer think DMB as TV. Consumers with less interest on digital products don't know well about DMB AD and then think DMB as cellular phone. In order to investigate perceptions of DMB as well as other digital instruments, we apply Proxscal analysis, Multidimensional Scaling technique at SPSS statistical package. At first step, subjects are presented 21 pairs of 7 digital instruments and evaluate similarity judgments on 7 point scale. And for each segment, their similarity judgments are averaged and similarity matrix is made. Secondly, Proxscal analysis of segment A and B are done. At third stage, get similarity judgment between DMB and other digital instruments after AD exposure. Lastly, similarity judgments of group A-1, A-2, B-1, and B-2 are named as 'after DMB' and put them into matrix made at the first stage. Then apply Proxscal analysis on these matrixes and check the positional difference of DMB and after DMB. The results show that map of segment A, who perceives DMB similar as TV, shows that DMB position closer to TV than to Cellular phone as expected. Map of segment B, who perceive DMB similar as cellular phone shows that DMB position closer to Cellular phone than to TV as expected. Stress value and R-square is acceptable. And, change results after stimuli, manipulated Advertising show that AD makes DMB perception bent toward Cellular phone when Cellular phone-like AD is exposed, and that DMB positioning move towards Car-TV which is more personalized one when TV-like AD is exposed. It is true for both segment, A and B, consistently. Furthermore, the paper apply correspondence analysis to the same data and find almost the same results. The paper answers two main research questions. The first one is that perception about a new product is made mainly from prior experience. And the second one is that AD is effective in changing and enforcing perception. In addition to above, we extend perception change to purchase intention. Purchase intention is high when AD enforces original perception. AD that shows DMB like TV makes worst intention. This paper has limitations and issues to be pursed in near future. Methodologically, current methodology can't provide statistical test on the perceptual change, since classical MDS models, like Proxscal and correspondence analysis are not probability models. So, a new probability MDS model for testing hypothesis about configuration needs to be developed. Next, advertising message needs to be developed more rigorously from theoretical and managerial perspective. Also experimental procedure could be improved for more realistic data collection. For example, web-based experiment and real product stimuli and multimedia presentation could be employed. Or, one can display products together in simulated shop. In addition, demand and social desirability threats of internal validity could influence on the results. In order to handle the threats, results of the model-intended advertising and other "pseudo" advertising could be compared. Furthermore, one can try various level of innovativeness in order to check whether it make any different results (cf. Moon 2006). In addition, if one can create hypothetical product that is really innovative and new for research, it helps to make a vacant impression status and then to study how to form impression in more rigorous way.