• Science of Emotion and Sensibility
• /
• v.13 no.1
• /
• pp.47-60
• /
• 2010

Most of the researches about classification usually have used kNN(k-Nearest Neighbor), SVM(Support Vector Machine), which are known as learn-based model, and Bayesian classifier, NNA(Neural Network Algorithm), which are known as statistics-based methods. However, there are some limitations of space and time when classifying so many web pages in recent internet. Moreover, most studies of classification are using uni-gram feature representation which is not good to represent real meaning of words. In case of Korean web page classification, there are some problems because of korean words property that the words have multiple meanings(polysemy). For these reasons, LSA(Latent Semantic Analysis) is proposed to classify well in these environment(large data set and words' polysemy). LSA uses SVD(Singular Value Decomposition) which decomposes the original term-document matrix to three different matrices and reduces their dimension. From this SVD's work, it is possible to create new low-level semantic space for representing vectors, which can make classification efficient and analyze latent meaning of words or document(or web pages). Although LSA is good at classification, it has some drawbacks in classification. As SVD reduces dimensions of matrix and creates new semantic space, it doesn't consider which dimensions discriminate vectors well but it does consider which dimensions represent vectors well. It is a reason why LSA doesn't improve performance of classification as expectation. In this paper, we propose new LSA which selects optimal dimensions to discriminate and represent vectors well as minimizing drawbacks and improving performance. This method that we propose shows better and more stable performance than other LSAs' in low-dimension space. In addition, we derive more improvement in classification as creating and selecting features by reducing stopwords and weighting specific values to them statistically.

• Journal of Intelligence and Information Systems
• /
• v.25 no.1
• /
• pp.163-177
• /
• 2019

As smartphones are getting widely used, human activity recognition (HAR) tasks for recognizing personal activities of smartphone users with multimodal data have been actively studied recently. The research area is expanding from the recognition of the simple body movement of an individual user to the recognition of low-level behavior and high-level behavior. However, HAR tasks for recognizing interaction behavior with other people, such as whether the user is accompanying or communicating with someone else, have gotten less attention so far. And previous research for recognizing interaction behavior has usually depended on audio, Bluetooth, and Wi-Fi sensors, which are vulnerable to privacy issues and require much time to collect enough data. Whereas physical sensors including accelerometer, magnetic field and gyroscope sensors are less vulnerable to privacy issues and can collect a large amount of data within a short time. In this paper, a method for detecting accompanying status based on deep learning model by only using multimodal physical sensor data, such as an accelerometer, magnetic field and gyroscope, was proposed. The accompanying status was defined as a redefinition of a part of the user interaction behavior, including whether the user is accompanying with an acquaintance at a close distance and the user is actively communicating with the acquaintance. A framework based on convolutional neural networks (CNN) and long short-term memory (LSTM) recurrent networks for classifying accompanying and conversation was proposed. First, a data preprocessing method which consists of time synchronization of multimodal data from different physical sensors, data normalization and sequence data generation was introduced. We applied the nearest interpolation to synchronize the time of collected data from different sensors. Normalization was performed for each x, y, z axis value of the sensor data, and the sequence data was generated according to the sliding window method. Then, the sequence data became the input for CNN, where feature maps representing local dependencies of the original sequence are extracted. The CNN consisted of 3 convolutional layers and did not have a pooling layer to maintain the temporal information of the sequence data. Next, LSTM recurrent networks received the feature maps, learned long-term dependencies from them and extracted features. The LSTM recurrent networks consisted of two layers, each with 128 cells. Finally, the extracted features were used for classification by softmax classifier. The loss function of the model was cross entropy function and the weights of the model were randomly initialized on a normal distribution with an average of 0 and a standard deviation of 0.1. The model was trained using adaptive moment estimation (ADAM) optimization algorithm and the mini batch size was set to 128. We applied dropout to input values of the LSTM recurrent networks to prevent overfitting. The initial learning rate was set to 0.001, and it decreased exponentially by 0.99 at the end of each epoch training. An Android smartphone application was developed and released to collect data. We collected smartphone data for a total of 18 subjects. Using the data, the model classified accompanying and conversation by 98.74% and 98.83% accuracy each. Both the F1 score and accuracy of the model were higher than the F1 score and accuracy of the majority vote classifier, support vector machine, and deep recurrent neural network. In the future research, we will focus on more rigorous multimodal sensor data synchronization methods that minimize the time stamp differences. In addition, we will further study transfer learning method that enables transfer of trained models tailored to the training data to the evaluation data that follows a different distribution. It is expected that a model capable of exhibiting robust recognition performance against changes in data that is not considered in the model learning stage will be obtained.

• Journal of Broadcast Engineering
• /
• v.19 no.5
• /
• pp.578-589
• /
• 2014

Window image is displayed through a monitor screen when we execute the application programs on the computer. This includes webpage, video player and a number of applications. The webpage delivers a variety of information by various types in comparison with other application. Unlike a natural image captured from a camera, the window image like a webpage includes diverse components such as text, logo, icon, subimage and so on. Each component delivers various types of information to users. However, the components with different characteristic need to be divided locally, because text and image are served by various type. In this paper, we divide window images into many sub blocks, and classify each divided region into background, text and subimage. The detected subimages can be applied into 2D-to-3D conversion, image retrieval, image browsing and so forth. There are many subimage classification methods. In this paper, we utilize AdaBoost for verifying that the machine learning-based algorithm can be efficient for subimage detection. In the experiment, we showed that the subimage detection ratio is 93.4 % and false alarm is 13 %.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.26 no.1
• /
• pp.187-196
• /
• 2016

Android applications are inherently vulnerable to a repackaging attack such that malicious codes are easily inserted into an application and then resigned by the attacker. These days, it occurs often that such private or individual information is leaked. In principle, all Android applications are composed of user defined methods and APIs. As well as accessing to resources on platform, APIs play a role as a practical functional feature, and user defined methods play a role as a feature by using APIs. In this paper we propose a scheme to analyze sensitive APIs mostly used in malicious applications in terms of how malicious applications operate and which API they use. Based on the characteristics of target APIs, we accumulate the knowledge on such APIs using a machine learning scheme based on Naive Bayes algorithm. Resulting from the learned results, we are able to provide fine-grained numeric score on the degree of vulnerabilities of mobile applications. In doing so, we expect the proposed scheme will help mobile application developers identify the security level of applications in advance.

• Steel and Composite Structures
• /
• v.34 no.5
• /
• pp.743-767
• /
• 2020

Due to the impressive flexural performance, enhanced compressive strength and more constrained crack propagation, Fibre-reinforced concrete (FRC) have been widely employed in the construction application. Majority of experimental studies have focused on the seismic behavior of FRC columns. Based on the valid experimental data obtained from the previous studies, the current study has evaluated the seismic response and compressive strength of FRC rectangular columns while following hybrid metaheuristic techniques. Due to the non-linearity of seismic data, Adaptive neuro-fuzzy inference system (ANFIS) has been incorporated with metaheuristic algorithms. 317 different datasets from FRC column tests has been applied as one database in order to determine the most influential factor on the ultimate strengths of FRC rectangular columns subjected to the simulated seismic loading. ANFIS has been used with the incorporation of Particle Swarm Optimization (PSO) and Genetic algorithm (GA). For the analysis of the attained results, Extreme learning machine (ELM) as an authentic prediction method has been concurrently used. The variable selection procedure is to choose the most dominant parameters affecting the ultimate strengths of FRC rectangular columns subjected to simulated seismic loading. Accordingly, the results have shown that ANFIS-PSO has successfully predicted the seismic lateral load with R² = 0.857 and 0.902 for the test and train phase, respectively, nominated as the lateral load prediction estimator. On the other hand, in case of compressive strength prediction, ELM is to predict the compressive strength with R² = 0.657 and 0.862 for test and train phase, respectively. The results have shown that the seismic lateral force trend is more predictable than the compressive strength of FRC rectangular columns, in which the best results belong to the lateral force prediction. Compressive strength prediction has illustrated a significant deviation above 40 Mpa which could be related to the considerable non-linearity and possible empirical shortcomings. Finally, employing ANFIS-GA and ANFIS-PSO techniques to evaluate the seismic response of FRC are a promising reliable approach to be replaced for high cost and time-consuming experimental tests.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.21 no.3
• /
• pp.414-419
• /
• 2020

An artificial intelligence-based cable tension estimation model was developed to expand the utilization of data obtained from cable accelerometers of cable-stayed bridges. The model was based on an algorithm for selecting the natural frequency in the tension estimation process based on the vibration method and an applied artificial neural network (ANN). The training data of the ANN was composed after converting the cable acceleration data into the frequency, and machine learning was carried out using the characteristics with a pattern on the natural frequency. When developing the training data, the frequencies with various amplitudes can be used to represent the frequencies of multiple shapes to improve the selection performance for natural frequencies. The performance of the model was estimated by comparing it with the control criteria of the tension estimated by an expert. As a result of the verification using 139 frequencies obtained from the cable accelerometer as the input, the natural frequency was determined to be similar to the real criteria and the estimated tension of the cable by the natural frequency was 96.4% of the criteria.

• KIPS Transactions on Software and Data Engineering
• /
• v.6 no.11
• /
• pp.527-536
• /
• 2017

With the recent development of smart grid industry, the necessity for efficient EMS(Energy Management System) has been increased. In particular, in order to reduce electric load and energy cost, sophisticated electric load forecasting and efficient smart grid operation strategy are required. In this paper, for more accurate electric load forecasting, we extend the data collected at demand time into high time resolution and construct an artificial neural network-based forecasting model appropriate for the high time resolution data. Furthermore, to improve the accuracy of electric load forecasting, time series data of sequence form are transformed into continuous data of two-dimensional space to solve that problem that machine learning methods cannot reflect the periodicity of time series data. In addition, to consider external factors such as temperature and humidity in accordance with the time resolution, we estimate their value at the time resolution using linear interpolation method. Finally, we apply the PCA(Principal Component Analysis) algorithm to the feature vector composed of external factors to remove data which have little correlation with the power data. Finally, we perform the evaluation of our model through 5-fold cross-validation. The results show that forecasting based on higher time resolution improve the accuracy and the best error rate of 3.71% was achieved at the 3-min resolution.

• KIPS Transactions on Software and Data Engineering
• /
• v.9 no.10
• /
• pp.317-322
• /
• 2020

Due to the development of artificial intelligence, it is used as a disaster response support system in the field of disaster. Disasters can occur anywhere, anytime. In the event of a disaster, there are four types of reports: fire, rescue, emergency, and other call. Disaster response according to the 119 call also responds differently depending on the type and situation. In this paper, 1280 data set of 119 calls were tested with 3 classes of SVM, NB, k-NN, DT, SGD, and RF situation classification algorithms using a training data set. Classification performance showed the highest performance of 92% and minimum of 77%. In the future, it is necessary to secure an effective data set by disaster in various fields to study disaster response.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.60 no.3
• /
• pp.639-647
• /
• 2011

In this paper, we introduce a design methodology of data-centroid Radial Basis Function neural networks with extended polynomial function. The two underlying design mechanisms of such networks involve K-means clustering method and Particle Swarm Optimization(PSO). The proposed algorithm is based on K-means clustering method for efficient processing of data and the optimization of model was carried out using PSO. In this paper, as the connection weight of RBF neural networks, we are able to use four types of polynomials such as simplified, linear, quadratic, and modified quadratic. Using K-means clustering, the center values of Gaussian function as activation function are selected. And the PSO-based RBF neural networks results in a structurally optimized structure and comes with a higher level of flexibility than the one encountered in the conventional RBF neural networks. The PSO-based design procedure being applied at each node of RBF neural networks leads to the selection of preferred parameters with specific local characteristics (such as the number of input variables, a specific set of input variables, and the distribution constant value in activation function) available within the RBF neural networks. To evaluate the performance of the proposed data-centroid RBF neural network with extended polynomial function, the model is experimented with using the nonlinear process data(2-Dimensional synthetic data and Mackey-Glass time series process data) and the Machine Learning dataset(NOx emission process data in gas turbine plant, Automobile Miles per Gallon(MPG) data, and Boston housing data). For the characteristic analysis of the given entire dataset with non-linearity as well as the efficient construction and evaluation of the dynamic network model, the partition of the given entire dataset distinguishes between two cases of Division I(training dataset and testing dataset) and Division II(training dataset, validation dataset, and testing dataset). A comparative analysis shows that the proposed RBF neural networks produces model with higher accuracy as well as more superb predictive capability than other intelligent models presented previously.

• KIISE Transactions on Computing Practices
• /
• v.21 no.11
• /
• pp.681-688
• /
• 2015

The k nearest neighbor (k-NN) graph construction is an important operation with many web-related applications, including collaborative filtering, similarity search, and many others in data mining and machine learning. Despite its many elegant properties, the brute force k-NN graph construction method has a computational complexity of $O(n^2)$, which is prohibitive for large scale data sets. Thus, (Key, Value)-based distributed framework, MapReduce, is gaining increasingly widespread use in Locality Sensitive Hashing which is efficient for high-dimension and sparse data. Based on the two-stage strategy, we engage the locality sensitive hashing technique to divide users into small subsets, and then calculate similarity between pairs in the small subsets using a brute force method on MapReduce. Specifically, generating a candidate group stage is important since brute-force calculation is performed in the following step. However, existing methods do not prevent large candidate groups. In this paper, we proposed an efficient algorithm for approximate k-NN graph construction by regrouping candidate groups. Experimental results show that our approach is more effective than existing methods in terms of graph accuracy and scan rate.