• Journal of Intelligence and Information Systems
• /
• v.23 no.3
• /
• pp.155-175
• /
• 2017

As social data become into the spotlight, mainstream web search engines provide data indicate how many people searched specific keyword: Web Search Traffic data. Web search traffic information is collection of each crowd that search for specific keyword. In a various area, web search traffic can be used as one of useful variables that represent the attention of common users on specific interests. A lot of studies uses web search traffic data to nowcast or forecast social phenomenon such as epidemic prediction, consumer pattern analysis, product life cycle, financial invest modeling and so on. Also web search traffic data have begun to be applied to predict tourist inbound. Proper demand prediction is needed because tourism is high value-added industry as increasing employment and foreign exchange. Among those tourists, especially Chinese tourists: Youke is continuously growing nowadays, Youke has been largest tourist inbound of Korea tourism for many years and tourism profits per one Youke as well. It is important that research into proper demand prediction approaches of Youke in both public and private sector. Accurate tourism demands prediction is important to efficient decision making in a limited resource. This study suggests improved model that reflects latest issue of society by presented the attention from group of individual. Trip abroad is generally high-involvement activity so that potential tourists likely deep into searching for information about their own trip. Web search traffic data presents tourists' attention in the process of preparation their journey instantaneous and dynamic way. So that this study attempted select key words that potential Chinese tourists likely searched out internet. Baidu-Chinese biggest web search engine that share over 80%- provides users with accessing to web search traffic data. Qualitative interview with potential tourists helps us to understand the information search behavior before a trip and identify the keywords for this study. Selected key words of web search traffic are categorized by how much directly related to "Korean Tourism" in a three levels. Classifying categories helps to find out which keyword can explain Youke inbound demands from close one to far one as distance of category. Web search traffic data of each key words gathered by web crawler developed to crawling web search data onto Baidu Index. Using automatically gathered variable data, linear model is designed by multiple regression analysis for suitable for operational application of decision and policy making because of easiness to explanation about variables' effective relationship. After regression linear models have composed, comparing with model composed traditional variables and model additional input web search traffic data variables to traditional model has conducted by significance and R squared. after comparing performance of models, final model is composed. Final regression model has improved explanation and advantage of real-time immediacy and convenience than traditional model. Furthermore, this study demonstrates system intuitively visualized to general use -Youke Mining solution has several functions of tourist decision making including embed final regression model. Youke Mining solution has algorithm based on data science and well-designed simple interface. In the end this research suggests three significant meanings on theoretical, practical and political aspects. Theoretically, Youke Mining system and the model in this research are the first step on the Youke inbound prediction using interactive and instant variable: web search traffic information represents tourists' attention while prepare their trip. Baidu web search traffic data has more than 80% of web search engine market. Practically, Baidu data could represent attention of the potential tourists who prepare their own tour as real-time. Finally, in political way, designed Chinese tourist demands prediction model based on web search traffic can be used to tourism decision making for efficient managing of resource and optimizing opportunity for successful policy.

• Korean Journal of Agricultural Science
• /
• v.16 no.2
• /
• pp.212-224
• /
• 1989

This paper reports results of an investigation to determine properties of superplasticizered fly ash concrete. The mixture proportions of fly ash were 0, 10, 20 and 30%, by weight of cement, and superplasticizer was added as a percentage of fly ash, 0, 0.6, 12 and 1.8%. To investigate the effective use of the superplasticized fly ash concrete, the basic data were analyzed. The results obtained were summarized as follows : 1. The unit water content was decreased by 1%, 6% and increased by 2% to the ratio of addition of fly ash 10%, 20%, 30%, respectively, but in case of the superplasticized fly ash concrete, it was decreased by 3~16%, 4~14% and 10~17%, at 0.6, 12, and 1.8% dosage of superplasticizer, respectively. 2. In the properties of the fresh fly ash concrete, the slump loss was reduced with the ratio of replacement of fly ash increased, and with times went by. When using superplasticizer in fly ash substituting concrete, the fludity in the concrete was not decreased. 3. The compressive strength of fly ash concrete at early ages was lower than that of ordinary concrete. At the later age of 28 days, the compressive strength with 20% addition of fly ash was increased than that of ordinary concrete. In cased of 10%, 30% addition of fly ash, the compressive strength were reduced. From this, it was proved that the optimum amount of fly ash appears to be about 20%. The compressive strength at all ages of superplasticized fly ash concrete was significantly higher than that of fly ash concrete, with increasing fly ash content. 4. In case of the tensile strength, the effects of the increasing strength with the ages were similar to those of the compressive strtength, and at the later ages was seen a decreasing tendency of strengths. 5. The correlation between compressive and tensile strength of superplasticized fly ash concrete was highly significant. The multiple regression equations of compressive and tensile strength were obtained on a function of the mixture proportion of fly ash and the addition of superplasticizer. The relation between compressive and tensile strength is higher than for ordinary concrete. The strength ratio is 7~11, and it is higher than that of ordinary concrete, 8~10. 6. Bulk density was decreased by 1~3% compared with ordinary concrete with the mixture proportion of fly ash increased, 10~30%, and decreased by 1~2% with the superplasticizer added 0.6~1.8%.

• The Journal of Korean Academy of Orthopedic Manual Physical Therapy
• /
• v.7 no.1
• /
• pp.5-28
• /
• 2001

Objectives: The objective of this study was to investigate analysis of bone mineral density according to Women with low back pain women. Methods: The data were collected from women who visited Physical Examination Center of a Catholic university hospital located in Daegu. Questionnaires were completed by 50 women during the period from July 20, 2000 to January 12, 2001. The sample was divided into three groups(the normal group of 16 cases and the osteopenia group of 12cases and the osteoporosis group of 22 cases). Bone mineral density(BMD) of lumbar spine was measured using energy absorptiometry. Results: The bone mineral density of the lumbar spine decreased with aging. The bone mineral density of the lumbar spine decreased with the serum Calcium and Phosphorus and Alkaline phosphatase increased. The mean bone mineral density of the lumbar spine of healthy women in age(50~59) was 0.87g/$cm^2$, the lumbar spine of women with low back pain in age(50~59) was 0.77g/$cm^2$. In the multiple regression of risk factors to bone mineral density(BMD) of lumbar spine were correlated with age, marriage existence, exercise time, the loving food of taste, calcium, bone mineral density standard T scores(p<0.05). The experience for LBP increased as weight increased(Odds ratio=999.000). The experience for LBP increased as number of Exercise decreased(Odds ratio=999.000). The experience for LBP increased as menopause existence increased(Odds ratio=999.000). The experience for LBP increased as serum Calcium and Phosphorus increased (Odds ratio=999.000). however all four variables had significant no relationship. The correlation in variables in relation to low back pain and bone mineral density, age showed contra-correlation with low back pain existence, Alkaline phosphatase(p<0.01). Weight showed contra-correlation with body mass index(BMI)(p<0.01). Exercise time showed correlation with number of exercise(p<0.01). The loving food of taste showed contra-correlation with Alkaline phosphatase(p<0.05). Bone mineral density showed correlation with menopause existence(p<0.05). Conclusions: Results from this study indicated that a statistically significant association between bone mineral density of the lumbar spin and age, marriage existence, exercise time, the loving food of taste, calcium, bone mineral density standard T scores. In logistic regression test, there were no related variables. The combination of bone mineral density measurement and assessment of the bone turnover rate by measuring biochemical would be helpful for the treatment of patients with risks of osteoporosis. The more precise study for risk factors to osteoporosis is essential.

• Journal of Intelligence and Information Systems
• /
• v.24 no.1
• /
• pp.205-225
• /
• 2018

Convolutional Neural Network (ConvNet) is one class of the powerful Deep Neural Network that can analyze and learn hierarchies of visual features. Originally, first neural network (Neocognitron) was introduced in the 80s. At that time, the neural network was not broadly used in both industry and academic field by cause of large-scale dataset shortage and low computational power. However, after a few decades later in 2012, Krizhevsky made a breakthrough on ILSVRC-12 visual recognition competition using Convolutional Neural Network. That breakthrough revived people interest in the neural network. The success of Convolutional Neural Network is achieved with two main factors. First of them is the emergence of advanced hardware (GPUs) for sufficient parallel computation. Second is the availability of large-scale datasets such as ImageNet (ILSVRC) dataset for training. Unfortunately, many new domains are bottlenecked by these factors. For most domains, it is difficult and requires lots of effort to gather large-scale dataset to train a ConvNet. Moreover, even if we have a large-scale dataset, training ConvNet from scratch is required expensive resource and time-consuming. These two obstacles can be solved by using transfer learning. Transfer learning is a method for transferring the knowledge from a source domain to new domain. There are two major Transfer learning cases. First one is ConvNet as fixed feature extractor, and the second one is Fine-tune the ConvNet on a new dataset. In the first case, using pre-trained ConvNet (such as on ImageNet) to compute feed-forward activations of the image into the ConvNet and extract activation features from specific layers. In the second case, replacing and retraining the ConvNet classifier on the new dataset, then fine-tune the weights of the pre-trained network with the backpropagation. In this paper, we focus on using multiple ConvNet layers as a fixed feature extractor only. However, applying features with high dimensional complexity that is directly extracted from multiple ConvNet layers is still a challenging problem. We observe that features extracted from multiple ConvNet layers address the different characteristics of the image which means better representation could be obtained by finding the optimal combination of multiple ConvNet layers. Based on that observation, we propose to employ multiple ConvNet layer representations for transfer learning instead of a single ConvNet layer representation. Overall, our primary pipeline has three steps. Firstly, images from target task are given as input to ConvNet, then that image will be feed-forwarded into pre-trained AlexNet, and the activation features from three fully connected convolutional layers are extracted. Secondly, activation features of three ConvNet layers are concatenated to obtain multiple ConvNet layers representation because it will gain more information about an image. When three fully connected layer features concatenated, the occurring image representation would have 9192 (4096+4096+1000) dimension features. However, features extracted from multiple ConvNet layers are redundant and noisy since they are extracted from the same ConvNet. Thus, a third step, we will use Principal Component Analysis (PCA) to select salient features before the training phase. When salient features are obtained, the classifier can classify image more accurately, and the performance of transfer learning can be improved. To evaluate proposed method, experiments are conducted in three standard datasets (Caltech-256, VOC07, and SUN397) to compare multiple ConvNet layer representations against single ConvNet layer representation by using PCA for feature selection and dimension reduction. Our experiments demonstrated the importance of feature selection for multiple ConvNet layer representation. Moreover, our proposed approach achieved 75.6% accuracy compared to 73.9% accuracy achieved by FC7 layer on the Caltech-256 dataset, 73.1% accuracy compared to 69.2% accuracy achieved by FC8 layer on the VOC07 dataset, 52.2% accuracy compared to 48.7% accuracy achieved by FC7 layer on the SUN397 dataset. We also showed that our proposed approach achieved superior performance, 2.8%, 2.1% and 3.1% accuracy improvement on Caltech-256, VOC07, and SUN397 dataset respectively compare to existing work.