• Journal of Intelligence and Information Systems
• /
• v.17 no.1
• /
• pp.53-69
• /
• 2011

Vision and voice-based technologies are commonly utilized for human-robot interaction. But it is widely recognized that the performance of vision and voice-based interaction systems is deteriorated by a large margin in the real-world situations due to environmental and user variances. Human users need to be very cooperative to get reasonable performance, which significantly limits the usability of the vision and voice-based human-robot interaction technologies. As a result, touch screens are still the major medium of human-robot interaction for the real-world applications. To empower the usability of robots for various services, alternative interaction technologies should be developed to complement the problems of vision and voice-based technologies. In this paper, we propose the use of accelerometer-based gesture interface as one of the alternative technologies, because accelerometers are effective in detecting the movements of human body, while their performance is not limited by environmental contexts such as lighting conditions or camera's field-of-view. Moreover, accelerometers are widely available nowadays in many mobile devices. We tackle the problem of classifying acceleration signal patterns of 26 English alphabets, which is one of the essential repertoires for the realization of education services based on robots. Recognizing 26 English handwriting patterns based on accelerometers is a very difficult task to take over because of its large scale of pattern classes and the complexity of each pattern. The most difficult problem that has been undertaken which is similar to our problem was recognizing acceleration signal patterns of 10 handwritten digits. Most previous studies dealt with pattern sets of 8~10 simple and easily distinguishable gestures that are useful for controlling home appliances, computer applications, robots etc. Good features are essential for the success of pattern recognition. To promote the discriminative power upon complex English alphabet patterns, we extracted 'motion trajectories' out of input acceleration signal and used them as the main feature. Investigative experiments showed that classifiers based on trajectory performed 3%~5% better than those with raw features e.g. acceleration signal itself or statistical figures. To minimize the distortion of trajectories, we applied a simple but effective set of smoothing filters and band-pass filters. It is well known that acceleration patterns for the same gesture is very different among different performers. To tackle the problem, online incremental learning is applied for our system to make it adaptive to the users' distinctive motion properties. Our system is based on instance-based learning (IBL) where each training sample is memorized as a reference pattern. Brute-force incremental learning in IBL continuously accumulates reference patterns, which is a problem because it not only slows down the classification but also downgrades the recall performance. Regarding the latter phenomenon, we observed a tendency that as the number of reference patterns grows, some reference patterns contribute more to the false positive classification. Thus, we devised an algorithm for optimizing the reference pattern set based on the positive and negative contribution of each reference pattern. The algorithm is performed periodically to remove reference patterns that have a very low positive contribution or a high negative contribution. Experiments were performed on 6500 gesture patterns collected from 50 adults of 30~50 years old. Each alphabet was performed 5 times per participant using $Nintendo{(R)}$ $Wii^{TM}$ remote. Acceleration signal was sampled in 100hz on 3 axes. Mean recall rate for all the alphabets was 95.48%. Some alphabets recorded very low recall rate and exhibited very high pairwise confusion rate. Major confusion pairs are D(88%) and P(74%), I(81%) and U(75%), N(88%) and W(100%). Though W was recalled perfectly, it contributed much to the false positive classification of N. By comparison with major previous results from VTT (96% for 8 control gestures), CMU (97% for 10 control gestures) and Samsung Electronics(97% for 10 digits and a control gesture), we could find that the performance of our system is superior regarding the number of pattern classes and the complexity of patterns. Using our gesture interaction system, we conducted 2 case studies of robot-based edutainment services. The services were implemented on various robot platforms and mobile devices including $iPhone^{TM}$. The participating children exhibited improved concentration and active reaction on the service with our gesture interface. To prove the effectiveness of our gesture interface, a test was taken by the children after experiencing an English teaching service. The test result showed that those who played with the gesture interface-based robot content marked 10% better score than those with conventional teaching. We conclude that the accelerometer-based gesture interface is a promising technology for flourishing real-world robot-based services and content by complementing the limits of today's conventional interfaces e.g. touch screen, vision and voice.

• Journal of Intelligence and Information Systems
• /
• v.19 no.4
• /
• pp.123-132
• /
• 2013

As the smartphones are equipped with various sensors such as the accelerometer, GPS, gravity sensor, gyros, ambient light sensor, proximity sensor, and so on, there have been many research works on making use of these sensors to create valuable applications. Human activity recognition is one such application that is motivated by various welfare applications such as the support for the elderly, measurement of calorie consumption, analysis of lifestyles, analysis of exercise patterns, and so on. One of the challenges faced when using the smartphone sensors for activity recognition is that the number of sensors used should be minimized to save the battery power. When the number of sensors used are restricted, it is difficult to realize a highly accurate activity recognizer or a classifier because it is hard to distinguish between subtly different activities relying on only limited information. The difficulty gets especially severe when the number of different activity classes to be distinguished is very large. In this paper, we show that a fairly accurate classifier can be built that can distinguish ten different activities by using only a single sensor data, i.e., the smartphone accelerometer data. The approach that we take to dealing with this ten-class problem is to use the ensemble of nested dichotomy (END) method that transforms a multi-class problem into multiple two-class problems. END builds a committee of binary classifiers in a nested fashion using a binary tree. At the root of the binary tree, the set of all the classes are split into two subsets of classes by using a binary classifier. At a child node of the tree, a subset of classes is again split into two smaller subsets by using another binary classifier. Continuing in this way, we can obtain a binary tree where each leaf node contains a single class. This binary tree can be viewed as a nested dichotomy that can make multi-class predictions. Depending on how a set of classes are split into two subsets at each node, the final tree that we obtain can be different. Since there can be some classes that are correlated, a particular tree may perform better than the others. However, we can hardly identify the best tree without deep domain knowledge. The END method copes with this problem by building multiple dichotomy trees randomly during learning, and then combining the predictions made by each tree during classification. The END method is generally known to perform well even when the base learner is unable to model complex decision boundaries As the base classifier at each node of the dichotomy, we have used another ensemble classifier called the random forest. A random forest is built by repeatedly generating a decision tree each time with a different random subset of features using a bootstrap sample. By combining bagging with random feature subset selection, a random forest enjoys the advantage of having more diverse ensemble members than a simple bagging. As an overall result, our ensemble of nested dichotomy can actually be seen as a committee of committees of decision trees that can deal with a multi-class problem with high accuracy. The ten classes of activities that we distinguish in this paper are 'Sitting', 'Standing', 'Walking', 'Running', 'Walking Uphill', 'Walking Downhill', 'Running Uphill', 'Running Downhill', 'Falling', and 'Hobbling'. The features used for classifying these activities include not only the magnitude of acceleration vector at each time point but also the maximum, the minimum, and the standard deviation of vector magnitude within a time window of the last 2 seconds, etc. For experiments to compare the performance of END with those of other methods, the accelerometer data has been collected at every 0.1 second for 2 minutes for each activity from 5 volunteers. Among these 5,900 ($=5{\times}(60{\times}2-2)/0.1$) data collected for each activity (the data for the first 2 seconds are trashed because they do not have time window data), 4,700 have been used for training and the rest for testing. Although 'Walking Uphill' is often confused with some other similar activities, END has been found to classify all of the ten activities with a fairly high accuracy of 98.4%. On the other hand, the accuracies achieved by a decision tree, a k-nearest neighbor, and a one-versus-rest support vector machine have been observed as 97.6%, 96.5%, and 97.6%, respectively.

• Journal of Intelligence and Information Systems
• /
• v.26 no.2
• /
• pp.1-25
• /
• 2020

In this paper, we suggest an application system architecture which provides accurate, fast and efficient automatic gasometer reading function. The system captures gasometer image using mobile device camera, transmits the image to a cloud server on top of private LTE network, and analyzes the image to extract character information of device ID and gas usage amount by selective optical character recognition based on deep learning technology. In general, there are many types of character in an image and optical character recognition technology extracts all character information in an image. But some applications need to ignore non-of-interest types of character and only have to focus on some specific types of characters. For an example of the application, automatic gasometer reading system only need to extract device ID and gas usage amount character information from gasometer images to send bill to users. Non-of-interest character strings, such as device type, manufacturer, manufacturing date, specification and etc., are not valuable information to the application. Thus, the application have to analyze point of interest region and specific types of characters to extract valuable information only. We adopted CNN (Convolutional Neural Network) based object detection and CRNN (Convolutional Recurrent Neural Network) technology for selective optical character recognition which only analyze point of interest region for selective character information extraction. We build up 3 neural networks for the application system. The first is a convolutional neural network which detects point of interest region of gas usage amount and device ID information character strings, the second is another convolutional neural network which transforms spatial information of point of interest region to spatial sequential feature vectors, and the third is bi-directional long short term memory network which converts spatial sequential information to character strings using time-series analysis mapping from feature vectors to character strings. In this research, point of interest character strings are device ID and gas usage amount. Device ID consists of 12 arabic character strings and gas usage amount consists of 4 ~ 5 arabic character strings. All system components are implemented in Amazon Web Service Cloud with Intel Zeon E5-2686 v4 CPU and NVidia TESLA V100 GPU. The system architecture adopts master-lave processing structure for efficient and fast parallel processing coping with about 700,000 requests per day. Mobile device captures gasometer image and transmits to master process in AWS cloud. Master process runs on Intel Zeon CPU and pushes reading request from mobile device to an input queue with FIFO (First In First Out) structure. Slave process consists of 3 types of deep neural networks which conduct character recognition process and runs on NVidia GPU module. Slave process is always polling the input queue to get recognition request. If there are some requests from master process in the input queue, slave process converts the image in the input queue to device ID character string, gas usage amount character string and position information of the strings, returns the information to output queue, and switch to idle mode to poll the input queue. Master process gets final information form the output queue and delivers the information to the mobile device. We used total 27,120 gasometer images for training, validation and testing of 3 types of deep neural network. 22,985 images were used for training and validation, 4,135 images were used for testing. We randomly splitted 22,985 images with 8:2 ratio for training and validation respectively for each training epoch. 4,135 test image were categorized into 5 types (Normal, noise, reflex, scale and slant). Normal data is clean image data, noise means image with noise signal, relfex means image with light reflection in gasometer region, scale means images with small object size due to long-distance capturing and slant means images which is not horizontally flat. Final character string recognition accuracies for device ID and gas usage amount of normal data are 0.960 and 0.864 respectively.

• Journal of Intelligence and Information Systems
• /
• v.22 no.3
• /
• pp.71-89
• /
• 2016

Sentiment analysis is used for identifying emotions or sentiments embedded in the user generated data such as customer reviews from blogs, social network services, and so on. Various research fields such as computer science and business management can take advantage of this feature to analyze customer-generated opinions. In previous studies, the star rating of a review is regarded as the same as sentiment embedded in the text. However, it does not always correspond to the sentiment polarity. Due to this supposition, previous studies have some limitations in their accuracy. To solve this issue, the present study uses a supervised sentiment classification model to measure a more accurate sentiment polarity. This study aims to propose an advanced sentiment classifier and to discover the correlation between movie reviews and box-office success. The advanced sentiment classifier is based on two supervised machine learning techniques, the Support Vector Machines (SVM) and Feedforward Neural Network (FNN). The sentiment scores of the movie reviews are measured by the sentiment classifier and are analyzed by statistical correlations between movie reviews and box-office success. Movie reviews are collected along with a star-rate. The dataset used in this study consists of 1,258,538 reviews from 175 films gathered from Naver Movie website (movie.naver.com). The results show that the proposed sentiment classifier outperforms Naive Bayes (NB) classifier as its accuracy is about 6% higher than NB. Furthermore, the results indicate that there are positive correlations between the star-rate and the number of audiences, which can be regarded as the box-office success of a movie. The study also shows that there is the mild, positive correlation between the sentiment scores estimated by the classifier and the number of audiences. To verify the applicability of the sentiment scores, an independent sample t-test was conducted. For this, the movies were divided into two groups using the average of sentiment scores. The two groups are significantly different in terms of the star-rated scores.

• Korean Journal of Health Education and Promotion
• /
• v.18 no.2
• /
• pp.45-63
• /
• 2001

The objective of this study is to give basic materials for selection and improvement of contents of safety education, which is substantially helpful to elementary students, by analysis of contents of safety education in some subjects and assessment of the needs of elementary students for safety education. For this purpose, this study was analyzed the contents of safety education in five subjects for elementary school and conducted the survey of 883 students in some elementary schools in Seoul from April 7 to 22, 2000. The results were as follows; 1. As a result of analysis of the proportion of contents regarding safety-related education in some subjects, Physical Education occupied the highest proportion (14.09％), and that was followed by Practical Subject (9.55％) and Moral Education (9.34％). However, the proportions in Social Study and Natural Science were very low, 1.85％ and 1.31％ each. In total lines of these five subjects, the numbers of line regarding safety education was contained by 5.78％. 2. Analyzing the proportion of domains of safety education in five textbooks, the Meaning of Safety and Basic Principles occupied the highest portion (29.5％), and that was followed by the Home Safety (24.0％), the Safety in School (17.1％), and the Play and Leisure Safety (14.0％). The Coping with Accidents and First Aid, the Safety from Fire and Explosion, and the Traffic Safety occupied relatively low portion, 6.9％, 5.7％, and 2.8％ each. 3. As a result of analysis of the proportion of the safety education domain in each subject, the Meaning of Safety and Basic Principles occupied the highest portion (23.6％) in Moral Education, the Home Safety (12.7％) in Practical Subject, and the Play and Leisure Safety (10.9％) in Physical Education. 4. Most of the participants in this survey experienced the Home Accidents (71.1％). And also, they experienced the Play and Leisure Accidents (57.9％), the Accidents in School (49.7％), the Traffic Accidents (45.3％), and the Fire and Explosion Accidents (24.7％) in order. 5. In the average proportion of the needs of participants for safety education in each domain, the Coping with Accidents and First Aid has the highest point (4.05). And, that was followed by the Home safety (3.79), the Safety from Fire and Explosion (3.73), the Meaning of Safety and Basic Principles (3.65), the Play and Leisure Safety (3.50), the Safety in School (3.37), and the Traffic Safety (3.35). The average proportion of the needs for safety education of total domains was 3.66. 6. In the needs for safety education regarding the feature of participants, it showed higher scores in female students than male ones (p〈0.001), in lower grader than higher grader (p〈0.05), and in the students born to wealth than those born poor (p〈0.05). Also, the children who recognize the necessity of safety education showed higher scores of the needs for safety education (p〈0.001). And it also showed the same results of high score to the children whose parents did the safety education (p〈0.00l) and to the children and their parents who have the higher degree of practicing safety (p〈0.001), and these differences were statistically significant. 7. In the extent of preference for methods of safety education, it showed high score to the Field Learning, followed by the Audio- Visual Education, the Discussion, and the Instruction of teacher. In the extent of preference for subjects regarding the contents of safety education by each domain, it showed high score to the subject of Safety for 4 domains - the Meaning of Safety and Basic Principles, the Traffic Safety, the Safety from Fire and Explosion, and the Coping with Accidents and First Aid. And also, they preferred Moral Education for 2 domains - the Home safety and the Safety in School, and Physical Education for a domain of the Play and Leisure Safety. 8. While 27 of 36 detail items was contained the contents of safety education, the proportion of needs of participants for safety education showed more than average 3.00 score in 34 of 36 detail items. However, none of 9 detail items was included in five textbooks. Also, 2 detail items - the Coping with Disasters and the Safety from Poisoning - were included together 2 parts; One part had the higher ranked 7 items acquired by analysis of the needs, and the other had the higher ranked 7 items acquired by analysis of the contents. But, except those 2 items, none of items were matched with each part.

• Journal of Intelligence and Information Systems
• /
• v.18 no.4
• /
• pp.1-17
• /
• 2012

Personalized smart devices such as smartphones and smart pads are widely used. Unlike traditional feature phones, theses smart devices allow users to choose a variety of functions, which support not only daily experiences but also business operations. Actually, there exist a huge number of applications accessible by smart device users in online and mobile application markets. Users can choose apps that fit their own tastes and needs, which is impossible for conventional phone users. With the increase in app demand, the tastes and needs of app users are becoming more diverse. To meet these requirements, numerous apps with diverse functions are being released on the market, which leads to fierce competition. Unlike offline markets, online markets have a limitation in that purchasing decisions should be made without experiencing the items. Therefore, online customers rely more on item-related information that can be seen on the item page in which online markets commonly provide details about each item. Customers can feel confident about the quality of an item through the online information and decide whether to purchase it. The same is true of online app markets. To win the sales competition against other apps that perform similar functions, app developers need to focus on writing app descriptions to attract the attention of customers. If we can measure the effect of app descriptions on sales without regard to the app's price and quality, app descriptions that facilitate the sale of apps can be identified. This study intends to provide such a quantitative result for app developers who want to promote the sales of their apps. For this purpose, we collected app details including the descriptions written in Korean from one of the largest app markets in Korea, and then extracted keywords from the descriptions. Next, the impact of the keywords on sales performance was measured through our econometric model. Through this analysis, we were able to analyze the impact of each keyword itself, apart from that of the design or quality. The keywords, comprised of the attribute and evaluation of each app, are extracted by a morpheme analyzer. Our model with the keywords as its input variables was established to analyze their impact on sales performance. A regression analysis was conducted for each category in which apps are included. This analysis was required because we found the keywords, which are emphasized in app descriptions, different category-by-category. The analysis conducted not only for free apps but also for paid apps showed which keywords have more impact on sales performance for each type of app. In the analysis of paid apps in the education category, keywords such as 'search+easy' and 'words+abundant' showed higher effectiveness. In the same category, free apps whose keywords emphasize the quality of apps showed higher sales performance. One interesting fact is that keywords describing not only the app but also the need for the app have asignificant impact. Language learning apps, regardless of whether they are sold free or paid, showed higher sales performance by including the keywords 'foreign language study+important'. This result shows that motivation for the purchase affected sales. While item reviews are widely researched in online markets, item descriptions are not very actively studied. In the case of the mobile app markets, newly introduced apps may not have many item reviews because of the low quantity sold. In such cases, item descriptions can be regarded more important when customers make a decision about purchasing items. This study is the first trial to quantitatively analyze the relationship between an item description and its impact on sales performance. The results show that our research framework successfully provides a list of the most effective sales key terms with the estimates of their effectiveness. Although this study is performed for a specified type of item (i.e., mobile apps), our model can be applied to almost all of the items traded in online markets.

• Asia pacific journal of information systems
• /
• v.19 no.2
• /
• pp.139-155
• /
• 2009

For the last few decades, many studies have tried to explore and unveil venture companies' success factors and unique features in order to identify the sources of such companies' competitive advantages over their rivals. Such venture companies have shown tendency to give high returns for investors generally making the best use of information technology. For this reason, many venture companies are keen on attracting avid investors' attention. Investors generally make their investment decisions by carefully examining the evaluation criteria of the alternatives. To them, credit rating information provided by international rating agencies, such as Standard and Poor's, Moody's and Fitch is crucial source as to such pivotal concerns as companies stability, growth, and risk status. But these types of information are generated only for the companies issuing corporate bonds, not venture companies. Therefore, this study proposes a method for evaluating venture businesses by presenting our recent empirical results using financial data of Korean venture companies listed on KOSDAQ in Korea exchange. In addition, this paper used multi-class SVM for the prediction of DEA-based efficiency rating for venture businesses, which was derived from our proposed method. Our approach sheds light on ways to locate efficient companies generating high level of profits. Above all, in determining effective ways to evaluate a venture firm's efficiency, it is important to understand the major contributing factors of such efficiency. Therefore, this paper is constructed on the basis of following two ideas to classify which companies are more efficient venture companies: i) making DEA based multi-class rating for sample companies and ii) developing multi-class SVM-based efficiency prediction model for classifying all companies. First, the Data Envelopment Analysis(DEA) is a non-parametric multiple input-output efficiency technique that measures the relative efficiency of decision making units(DMUs) using a linear programming based model. It is non-parametric because it requires no assumption on the shape or parameters of the underlying production function. DEA has been already widely applied for evaluating the relative efficiency of DMUs. Recently, a number of DEA based studies have evaluated the efficiency of various types of companies, such as internet companies and venture companies. It has been also applied to corporate credit ratings. In this study we utilized DEA for sorting venture companies by efficiency based ratings. The Support Vector Machine(SVM), on the other hand, is a popular technique for solving data classification problems. In this paper, we employed SVM to classify the efficiency ratings in IT venture companies according to the results of DEA. The SVM method was first developed by Vapnik (1995). As one of many machine learning techniques, SVM is based on a statistical theory. Thus far, the method has shown good performances especially in generalizing capacity in classification tasks, resulting in numerous applications in many areas of business, SVM is basically the algorithm that finds the maximum margin hyperplane, which is the maximum separation between classes. According to this method, support vectors are the closest to the maximum margin hyperplane. If it is impossible to classify, we can use the kernel function. In the case of nonlinear class boundaries, we can transform the inputs into a high-dimensional feature space, This is the original input space and is mapped into a high-dimensional dot-product space. Many studies applied SVM to the prediction of bankruptcy, the forecast a financial time series, and the problem of estimating credit rating, In this study we employed SVM for developing data mining-based efficiency prediction model. We used the Gaussian radial function as a kernel function of SVM. In multi-class SVM, we adopted one-against-one approach between binary classification method and two all-together methods, proposed by Weston and Watkins(1999) and Crammer and Singer(2000), respectively. In this research, we used corporate information of 154 companies listed on KOSDAQ market in Korea exchange. We obtained companies' financial information of 2005 from the KIS(Korea Information Service, Inc.). Using this data, we made multi-class rating with DEA efficiency and built multi-class prediction model based data mining. Among three manners of multi-classification, the hit ratio of the Weston and Watkins method is the best in the test data set. In multi classification problems as efficiency ratings of venture business, it is very useful for investors to know the class with errors, one class difference, when it is difficult to find out the accurate class in the actual market. So we presented accuracy results within 1-class errors, and the Weston and Watkins method showed 85.7% accuracy in our test samples. We conclude that the DEA based multi-class approach in venture business generates more information than the binary classification problem, notwithstanding its efficiency level. We believe this model can help investors in decision making as it provides a reliably tool to evaluate venture companies in the financial domain. For the future research, we perceive the need to enhance such areas as the variable selection process, the parameter selection of kernel function, the generalization, and the sample size of multi-class.

• Journal of the Korean Chemical Society
• /
• v.59 no.2
• /
• pp.164-178
• /
• 2015

The purpose of this study was to design a teaching and learning method conductive to the development of creative thinking skills and investigate its effects. It has been developed integrated mind map with feature of visualizing the divergent thinking to the aspects of Science (S), Technology (T) & Engineering (E), Arts (A), Mathematics (M). Integrated mind map can be divided into four types of STEAM type, STEA type, STEM type, STE type depending on the category of key words in the first branch. And Integrated mind map can be divided into three levels of guided, intermediate, open depending on the teacher's guide degree. And also integrated mind map activities were carried out in the form of group, class share as well as individual. This study was implemented during a semester and students in experiment group experienced individual-integrated mind map activity 10 times, group-integrated mind map activity 10 times, class share-integrated mind map activity 3 times. The results indicated that the experimental group presented statistically meaningful improvement in creative thinking skills (p<.05). And there was a statistically meaningful improvement in fluency, flexibility, originality as a sub-category of creative thinking skills(p <.05). Also creative thinking skills are not affected by the level of cognitive, academic performance, gender (p<.05). In conclusion, it was found that 'integrated mind map activity' improved student's creative thinking skills. There was no interaction effect about creative thinking skills between the teaching strategy and cognitive level, achivement, gender of those students.

• Journal of Korean Home Economics Education Association
• /
• v.20 no.4
• /
• pp.157-171
• /
• 2008

The year of 2007 Reformed Curriculum encourages various activity materials in the textbook facilitate students oriented self-help learning. The purpose of this paper is to find out how much the activity materials in housing area of middle school Technology and Home Economics are practiced in the class and why they are used or not used. The data were collected from 253 middle school teachers who had ever taught the housing unit in any of 6 textbooks. The analyses indicated that the most frequent teaching methode was lecture based on the textbook and internet data focused on the figures and contents of the individual textbook. The average rate of practicing the activity materials was differ by textbooks and the characteristics of the materials such as type of materials, feature of non sentence materials, and type of activity. The main two reasons to practice the activity materials were it's adequacy to class goals and application to everyday life. Low interests of students and shortage of time were the two main reasons why not used the materials. Textbook writers should consider these reasons as well as the characteristics of activity materials practiced in the class by the teachers in order to meet the goals of the reformed as well as current curricula.

• Journal of Intelligence and Information Systems
• /
• v.18 no.1
• /
• pp.39-57
• /
• 2012

Recently, due to the introduction of high-tech equipment in interactive exhibits, many people's attention has been concentrated on Interactive exhibits that can double the exhibition effect through the interaction with the audience. In addition, it is also possible to measure a variety of audience reaction in the interactive exhibition. Among various audience reactions, this research uses the change of the facial features that can be collected in an interactive exhibition space. This research develops an artificial neural network-based prediction model to predict the response of the audience by measuring the change of the facial features when the audience is given stimulation from the non-excited state. To present the emotion state of the audience, this research uses a Valence-Arousal model. So, this research suggests an overall framework composed of the following six steps. The first step is a step of collecting data for modeling. The data was collected from people participated in the 2012 Seoul DMC Culture Open, and the collected data was used for the experiments. The second step extracts 64 facial features from the collected data and compensates the facial feature values. The third step generates independent and dependent variables of an artificial neural network model. The fourth step extracts the independent variable that affects the dependent variable using the statistical technique. The fifth step builds an artificial neural network model and performs a learning process using train set and test set. Finally the last sixth step is to validate the prediction performance of artificial neural network model using the validation data set. The proposed model is compared with statistical predictive model to see whether it had better performance or not. As a result, although the data set in this experiment had much noise, the proposed model showed better results when the model was compared with multiple regression analysis model. If the prediction model of audience reaction was used in the real exhibition, it will be able to provide countermeasures and services appropriate to the audience's reaction viewing the exhibits. Specifically, if the arousal of audience about Exhibits is low, Action to increase arousal of the audience will be taken. For instance, we recommend the audience another preferred contents or using a light or sound to focus on these exhibits. In other words, when planning future exhibitions, planning the exhibition to satisfy various audience preferences would be possible. And it is expected to foster a personalized environment to concentrate on the exhibits. But, the proposed model in this research still shows the low prediction accuracy. The cause is in some parts as follows : First, the data covers diverse visitors of real exhibitions, so it was difficult to control the optimized experimental environment. So, the collected data has much noise, and it would results a lower accuracy. In further research, the data collection will be conducted in a more optimized experimental environment. The further research to increase the accuracy of the predictions of the model will be conducted. Second, using changes of facial expression only is thought to be not enough to extract audience emotions. If facial expression is combined with other responses, such as the sound, audience behavior, it would result a better result.