• Title/Summary/Keyword: machine learning framework

Search Result 250, Processing Time 0.024 seconds

A Hybrid SVM Classifier for Imbalanced Data Sets (불균형 데이터 집합의 분류를 위한 하이브리드 SVM 모델)

  • Lee, Jae Sik;Kwon, Jong Gu
    • Journal of Intelligence and Information Systems
    • /
    • v.19 no.2
    • /
    • pp.125-140
    • /
    • 2013
  • We call a data set in which the number of records belonging to a certain class far outnumbers the number of records belonging to the other class, 'imbalanced data set'. Most of the classification techniques perform poorly on imbalanced data sets. When we evaluate the performance of a certain classification technique, we need to measure not only 'accuracy' but also 'sensitivity' and 'specificity'. In a customer churn prediction problem, 'retention' records account for the majority class, and 'churn' records account for the minority class. Sensitivity measures the proportion of actual retentions which are correctly identified as such. Specificity measures the proportion of churns which are correctly identified as such. The poor performance of the classification techniques on imbalanced data sets is due to the low value of specificity. Many previous researches on imbalanced data sets employed 'oversampling' technique where members of the minority class are sampled more than those of the majority class in order to make a relatively balanced data set. When a classification model is constructed using this oversampled balanced data set, specificity can be improved but sensitivity will be decreased. In this research, we developed a hybrid model of support vector machine (SVM), artificial neural network (ANN) and decision tree, that improves specificity while maintaining sensitivity. We named this hybrid model 'hybrid SVM model.' The process of construction and prediction of our hybrid SVM model is as follows. By oversampling from the original imbalanced data set, a balanced data set is prepared. SVM_I model and ANN_I model are constructed using the imbalanced data set, and SVM_B model is constructed using the balanced data set. SVM_I model is superior in sensitivity and SVM_B model is superior in specificity. For a record on which both SVM_I model and SVM_B model make the same prediction, that prediction becomes the final solution. If they make different prediction, the final solution is determined by the discrimination rules obtained by ANN and decision tree. For a record on which SVM_I model and SVM_B model make different predictions, a decision tree model is constructed using ANN_I output value as input and actual retention or churn as target. We obtained the following two discrimination rules: 'IF ANN_I output value <0.285, THEN Final Solution = Retention' and 'IF ANN_I output value ${\geq}0.285$, THEN Final Solution = Churn.' The threshold 0.285 is the value optimized for the data used in this research. The result we present in this research is the structure or framework of our hybrid SVM model, not a specific threshold value such as 0.285. Therefore, the threshold value in the above discrimination rules can be changed to any value depending on the data. In order to evaluate the performance of our hybrid SVM model, we used the 'churn data set' in UCI Machine Learning Repository, that consists of 85% retention customers and 15% churn customers. Accuracy of the hybrid SVM model is 91.08% that is better than that of SVM_I model or SVM_B model. The points worth noticing here are its sensitivity, 95.02%, and specificity, 69.24%. The sensitivity of SVM_I model is 94.65%, and the specificity of SVM_B model is 67.00%. Therefore the hybrid SVM model developed in this research improves the specificity of SVM_B model while maintaining the sensitivity of SVM_I model.

Gridded Expansion of Forest Flux Observations and Mapping of Daily CO2 Absorption by the Forests in Korea Using Numerical Weather Prediction Data and Satellite Images (국지예보모델과 위성영상을 이용한 극상림 플럭스 관측의 공간연속면 확장 및 우리나라 산림의 일일 탄소흡수능 격자자료 산출)

  • Kim, Gunah;Cho, Jaeil;Kang, Minseok;Lee, Bora;Kim, Eun-Sook;Choi, Chuluong;Lee, Hanlim;Lee, Taeyun;Lee, Yangwon
    • Korean Journal of Remote Sensing
    • /
    • v.36 no.6_1
    • /
    • pp.1449-1463
    • /
    • 2020
  • As recent global warming and climate changes become more serious, the importance of CO2 absorption by forests is increasing to cope with the greenhouse gas issues. According to the UN Framework Convention on Climate Change, it is required to calculate national CO2 absorptions at the local level in a more scientific and rigorous manner. This paper presents the gridded expansion of forest flux observations and mapping of daily CO2 absorption by the forests in Korea using numerical weather prediction data and satellite images. To consider the sensitive daily changes of plant photosynthesis, we built a machine learning model to retrieve the daily RACA (reference amount of CO2 absorption) by referring to the climax forest in Gwangneung and adopted the NIFoS (National Institute of Forest Science) lookup table for the CO2 absorption by forest type and age to produce the daily AACA (actual amount of CO2 absorption) raster data with the spatial variation of the forests in Korea. In the experiment for the 1,095 days between Jan 1, 2013 and Dec 31, 2015, our RACA retrieval model showed high accuracy with a correlation coefficient of 0.948. To achieve the tier 3 daily statistics for AACA, long-term and detailed forest surveying should be combined with the model in the future.

Machine Learning Based Automated Source, Sink Categorization for Hybrid Approach of Privacy Leak Detection (머신러닝 기반의 자동화된 소스 싱크 분류 및 하이브리드 분석을 통한 개인정보 유출 탐지 방법)

  • Shim, Hyunseok;Jung, Souhwan
    • Journal of the Korea Institute of Information Security & Cryptology
    • /
    • v.30 no.4
    • /
    • pp.657-667
    • /
    • 2020
  • The Android framework allows apps to take full advantage of personal information through granting single permission, and does not determine whether the data being leaked is actual personal information. To solve these problems, we propose a tool with static/dynamic analysis. The tool analyzes the Source and Sink used by the target app, to provide users with information on what personal information it used. To achieve this, we extracted the Source and Sink through Control Flow Graph and make sure that it leaks the user's privacy when there is a Source-to-Sink flow. We also used the sensitive permission information provided by Google to obtain information from the sensitive API corresponding to Source and Sink. Finally, our dynamic analysis tool runs the app and hooks information from each sensitive API. In the hooked data, we got information about whether user's personal information is leaked through this app, and delivered to user. In this process, an automated Source/Sink classification model was applied to collect latest Source/Sink information, and the we categorized latest release version of Android(9.0) with 88.5% accuracy. We evaluated our tool on 2,802 APKs, and found 850 APKs that leak personal information.

Spark based Scalable RDFS Ontology Reasoning over Big Triples with Confidence Values (신뢰값 기반 대용량 트리플 처리를 위한 스파크 환경에서의 RDFS 온톨로지 추론)

  • Park, Hyun-Kyu;Lee, Wan-Gon;Jagvaral, Batselem;Park, Young-Tack
    • Journal of KIISE
    • /
    • v.43 no.1
    • /
    • pp.87-95
    • /
    • 2016
  • Recently, due to the development of the Internet and electronic devices, there has been an enormous increase in the amount of available knowledge and information. As this growth has proceeded, studies on large-scale ontological reasoning have been actively carried out. In general, a machine learning program or knowledge engineer measures and provides a degree of confidence for each triple in a large ontology. Yet, the collected ontology data contains specific uncertainty and reasoning such data can cause vagueness in reasoning results. In order to solve the uncertainty issue, we propose an RDFS reasoning approach that utilizes confidence values indicating degrees of uncertainty in the collected data. Unlike conventional reasoning approaches that have not taken into account data uncertainty, by using the in-memory based cluster computing framework Spark, our approach computes confidence values in the data inferred through RDFS-based reasoning by applying methods for uncertainty estimating. As a result, the computed confidence values represent the uncertainty in the inferred data. To evaluate our approach, ontology reasoning was carried out over the LUBM standard benchmark data set with addition arbitrary confidence values to ontology triples. Experimental results indicated that the proposed system is capable of running over the largest data set LUBM3000 in 1179 seconds inferring 350K triples.

An Improvement in K-NN Graph Construction using re-grouping with Locality Sensitive Hashing on MapReduce (MapReduce 환경에서 재그룹핑을 이용한 Locality Sensitive Hashing 기반의 K-Nearest Neighbor 그래프 생성 알고리즘의 개선)

  • Lee, Inhoe;Oh, Hyesung;Kim, Hyoung-Joo
    • 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.

Causal inference from nonrandomized data: key concepts and recent trends (비실험 자료로부터의 인과 추론: 핵심 개념과 최근 동향)

  • Choi, Young-Geun;Yu, Donghyeon
    • The Korean Journal of Applied Statistics
    • /
    • v.32 no.2
    • /
    • pp.173-185
    • /
    • 2019
  • Causal questions are prevalent in scientific research, for example, how effective a treatment was for preventing an infectious disease, how much a policy increased utility, or which advertisement would give the highest click rate for a given customer. Causal inference theory in statistics interprets those questions as inferring the effect of a given intervention (treatment or policy) in the data generating process. Causal inference has been used in medicine, public health, and economics; in addition, it has received recent attention as a tool for data-driven decision making processes. Many recent datasets are observational, rather than experimental, which makes the causal inference theory more complex. This review introduces key concepts and recent trends of statistical causal inference in observational studies. We first introduce the Neyman-Rubin's potential outcome framework to formularize from causal questions to average treatment effects as well as discuss popular methods to estimate treatment effects such as propensity score approaches and regression approaches. For recent trends, we briefly discuss (1) conditional (heterogeneous) treatment effects and machine learning-based approaches, (2) curse of dimensionality on the estimation of treatment effect and its remedies, and (3) Pearl's structural causal model to deal with more complex causal relationships and its connection to the Neyman-Rubin's potential outcome model.

A decision-centric impact assessment of operational performance of the Yongdam Dam, South Korea (용담댐 기존운영에 대한 의사결정중심 기후변화 영향 평가)

  • Kim, Daeha;Kim, Eunhee;Lee, Seung Cheol;Kim, Eunji;Shin, June
    • Journal of Korea Water Resources Association
    • /
    • v.55 no.3
    • /
    • pp.205-215
    • /
    • 2022
  • Amidst the global climate crisis, dam operation policies formulated under the stationary climate assumption could lead to unsatisfactory water management. In this work, we assessed status-quo performance of the Yongdam Dam in Korea under various climatic stresses in flood risk reduction and water supply reliability for 2021-2040. To this end, we employed a decision-centric framework equipped with a stochastic weather generator, a conceptual streamflow model, and a machine-learning reservoir operation rule. By imposing 294 climate perturbations to dam release simulations, we found that the current operation rule of the Yongdam dam could redundantly secure water storage, while inefficiently enhancing the supply reliability. On the other hand, flood risks were likely to increase substantially due to rising mean and variability of daily precipitation. Here, we argue that the current operation rules of the Yongdam Dam seem to be overly focused on securing water storage, and thus need to be adjusted to efficiently improve supply reliability and reduce flood risks in downstream areas.

Mapping Mammalian Species Richness Using a Machine Learning Algorithm (머신러닝 알고리즘을 이용한 포유류 종 풍부도 매핑 구축 연구)

  • Zhiying Jin;Dongkun Lee;Eunsub Kim;Jiyoung Choi;Yoonho Jeon
    • Journal of Environmental Impact Assessment
    • /
    • v.33 no.2
    • /
    • pp.53-63
    • /
    • 2024
  • Biodiversity holds significant importance within the framework of environmental impact assessment, being utilized in site selection for development, understanding the surrounding environment, and assessing the impact on species due to disturbances. The field of environmental impact assessment has seen substantial research exploring new technologies and models to evaluate and predict biodiversity more accurately. While current assessments rely on data from fieldwork and literature surveys to gauge species richness indices, limitations in spatial and temporal coverage underscore the need for high-resolution biodiversity assessments through species richness mapping. In this study, leveraging data from the 4th National Ecosystem Survey and environmental variables, we developed a species distribution model using Random Forest. This model yielded mapping results of 24 mammalian species' distribution, utilizing the species richness index to generate a 100-meter resolution map of species richness. The research findings exhibited a notably high predictive accuracy, with the species distribution model demonstrating an average AUC value of 0.82. In addition, the comparison with National Ecosystem Survey data reveals that the species richness distribution in the high-resolution species richness mapping results conforms to a normal distribution. Hence, it stands as highly reliable foundational data for environmental impact assessment. Such research and analytical outcomes could serve as pivotal new reference materials for future urban development projects, offering insights for biodiversity assessment and habitat preservation endeavors.

A Deep Learning Based Approach to Recognizing Accompanying Status of Smartphone Users Using Multimodal Data (스마트폰 다종 데이터를 활용한 딥러닝 기반의 사용자 동행 상태 인식)

  • Kim, Kilho;Choi, Sangwoo;Chae, Moon-jung;Park, Heewoong;Lee, Jaehong;Park, Jonghun
    • 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.

The Dynamics of CO2 Budget in Gwangneung Deciduous Old-growth Forest: Lessons from the 15 years of Monitoring (광릉 낙엽활엽수 노령림의 CO2 수지 역학: 15년 관측으로부터의 교훈)

  • Yang, Hyunyoung;Kang, Minseok;Kim, Joon;Ryu, Daun;Kim, Su-Jin;Chun, Jung-Hwa;Lim, Jong-Hwan;Park, Chan Woo;Yun, Soon Jin
    • Korean Journal of Agricultural and Forest Meteorology
    • /
    • v.23 no.4
    • /
    • pp.198-221
    • /
    • 2021
  • After large-scale reforestation in the 1960s and 1970s, forests in Korea have gradually been aging. Net ecosystem CO2 exchange of old-growth forests is theoretically near zero; however, it can be a CO2 sink or source depending on the intervention of disturbance or management. In this study, we report the CO2 budget dynamics of the Gwangneung deciduous old-growth forest (GDK) in Korea and examined the following two questions: (1) is the preserved GDK indeed CO2 neutral as theoretically known? and (2) can we explain the dynamics of CO2 budget by the common mechanisms reported in the literature? To answer, we analyzed the 15-year long CO2 flux data measured by eddy covariance technique along with other biometeorological data at the KoFlux GDK site from 2006 to 2020. The results showed that (1) GDK switched back-and-forth between sink and source of CO2 but averaged to be a week CO2 source (and turning to a moderate CO2 source for the recent five years) and (2) the interannual variability of solar radiation, growing season length, and leaf area index showed a positive correlation with that of gross primary production (GPP) (R2=0.32~0.45); whereas the interannual variability of both air and surface temperature was not significantly correlated with that of ecosystem respiration (RE). Furthermore, the machine learning-based model trained using the dataset of early monitoring period (first 10 years) failed to reproduce the observed interannual variations of GPP and RE for the recent five years. Biomass data analysis suggests that carbon emissions from coarse woody debris may have contributed partly to the conversion to a moderate CO2 source. To properly understand and interpret the long-term CO2 budget dynamics of GDK, new framework of analysis and modeling based on complex systems science is needed. Also, it is important to maintain the flux monitoring and data quality along with the monitoring of coarse woody debris and disturbances.