• 한국컴퓨터정보학회논문지
• /
• 제29권7호
• /
• pp.73-80
• /
• 2024

본 연구는 토픽 모델링과 장단기 기억(LSTM) 신경망을 결합하여 한국 종합주가지수(KOSPI) 예측의 정확도를 향상하는 방법을 제안한다. 본 논문에서는 LDA(Latent Dirichlet Allocation) 기법을 이용해 금융 뉴스 데이터에서 금리 인상 및 인하와 관련된 10개의 주요 주제를 추출하고, 추출된 주제를 과거 KOSPI 지수와 함께 LSTM 모델에 입력하여 KOSPI 지수를 예측하는 모델을 제안한다. 제안된 모델은 과거 KOSPI 지수를 LSTM 모델에 입력하여 시계열 예측 방법과 뉴스 데이터를 입력하여 토픽 모델링하는 방법을 결합하여 KOSPI 지수를 예측하는 특성을 가진다. 제안된 모델의 성능을 검증하기 위해, 본 논문에서는 LSTM의 입력 데이터의 종류에 따라 4개의 모델(LSTM_K 모델, LSTM_KNS 모델, LDA_K 모델, LDA_KNS 모델)을 설계하고 각 모델의 예측 성능을 제시하였다. 예측 성능을 비교한 결과, 금융 뉴스 주제 데이터와 과거 KOSPI 지수 데이터를 입력으로 하는 LSTM 모델(LDA_K 모델)이 가장 낮은 RMSE(Root Mean Square Error)를 기록하여 가장 좋은 예측 성능을 보였다.

• The Journal of Asian Finance, Economics and Business
• /
• 제9권4호
• /
• pp.83-92
• /
• 2022

Due to the uncertainty in the order of the integrated model, the SARIMA-LSTM model, SARIMA-SVR model, LSTM-SARIMA model, and SVR-SARIMA model are constructed respectively to determine the best-combined model for forecasting the China-Russia trade turnover. Meanwhile, the effect of the order of the combined models on the prediction results is analyzed. Using indicators such as MAPE and RMSE, we compare and evaluate the predictive effects of different models. The results show that the SARIMA-LSTM model combines the SARIMA model's short-term forecasting advantage with the LSTM model's long-term forecasting advantage, which has the highest forecast accuracy of all models and can accurately predict the trend of China-Russia trade turnover in the post-epidemic period. Furthermore, the SARIMA - LSTM model has a higher forecast accuracy than the LSTM-ARIMA model. Nevertheless, the SARIMA-SVR model's forecast accuracy is lower than the SVR-SARIMA model's. As a result, the combined models' order has no bearing on the predicting outcomes for the China-Russia trade turnover time series.

• Journal of Information Processing Systems
• /
• 제17권4호
• /
• pp.818-833
• /
• 2021

To address the problems of low precision rate, insufficient feature extraction, and poor contextual ability in existing text sentiment analysis methods, a mixed model account of a CNN-BiLSTM-TE (convolutional neural network, bidirectional long short-term memory, and topic extraction) model was proposed. First, Chinese text data was converted into vectors through the method of transfer learning by Word2Vec. Second, local features were extracted by the CNN model. Then, contextual information was extracted by the BiLSTM neural network and the emotional tendency was obtained using softmax. Finally, topics were extracted by the term frequency-inverse document frequency and K-means. Compared with the CNN, BiLSTM, and gate recurrent unit (GRU) models, the CNN-BiLSTM-TE model's F1-score was higher than other models by 0.0147, 0.006, and 0.0052, respectively. Then compared with CNN-LSTM, LSTM-CNN, and BiLSTM-CNN models, the F1-score was higher by 0.0071, 0.0038, and 0.0049, respectively. Experimental results showed that the CNN-BiLSTM-TE model can effectively improve various indicators in application. Lastly, performed scalability verification through a takeaway dataset, which has great value in practical applications.

• 한국수자원학회논문집
• /
• 제55권7호
• /
• pp.495-504
• /
• 2022

최근 인공지능의 발전으로 시계열 자료 분석에 효과적인 Long Short-Term Memory (LSTM) 모델이 댐유입량 예측의 정확도를 높이는 데 활용되고 있다. 본 연구에서는 그 중 LSTM의 성능을 더욱 향상할 수 있는 Sequence-to-Sequence (s2s) 구조에 Attention 기법을 LSTM 모델에 첨가하여 소양강댐 유역의 유입량을 예측하였다. 분석 데이터는 2013년부터 2020년까지의 유입량 시자료와 종관기상관측기온 및 강수량 자료를 학습, 검증, 평가로 나누어 훈련한 후, 모델의 성능 평가를 진행하였다. 분석 결과, LSTM-s2s 모델보다 attention까지 첨가한 모델이 일반적으로 더 좋은 성능을 보였으며, attention 첨가 모델이 첨두값도 더 잘 예측하는 모습을 보였다. 그리고 두 모델 모두 첨두값 발생 동안 유량 패턴을 잘 반영하였지만 세밀한 시간 단위 변화량에는 어려움이 있었다. 이를 통해 시간 단위 예측의 어려움에도 불구하고, LSTM-s2s에 attention까지 첨가한 모델이 기존 LSTM-s2s의 예측 성능을 향상할 수 있음을 알 수 있었다.

• International Journal of Computer Science & Network Security
• /
• 제22권4호
• /
• pp.119-130
• /
• 2022

Crime is a common social problem that affects the quality of life. As the number of crimes increases, it is necessary to build a model to predict the number of crimes that may occur in a given period, identify the characteristics of a person who may commit a particular crime, and identify places where a particular crime may occur. Data privacy is the main challenge that organizations face when building this type of predictive models. Federated learning (FL) is a promising approach that overcomes data security and privacy challenges, as it enables organizations to build a machine learning model based on distributed datasets without sharing raw data or violating data privacy. In this paper, a federated long short- term memory (LSTM) model is proposed and compared with a traditional LSTM model. Proposed model is developed using TensorFlow Federated (TFF) and the Keras API to predict the number of crimes. The proposed model is applied on the Boston crime dataset. The proposed model's parameters are fine tuned to obtain minimum loss and maximum accuracy. The proposed federated LSTM model is compared with the traditional LSTM model and found that the federated LSTM model achieved lower loss, better accuracy, and higher training time than the traditional LSTM model.

• 정보처리학회논문지:소프트웨어 및 데이터공학
• /
• 제10권11호
• /
• pp.491-500
• /
• 2021

본 연구에서는 MS Kinect v2 RGBD 카메라 기반의 Human-Skeleton Keypoints와 2-Stacked Bi-LSTM 모델을 이용하여 낙상 행위를 탐지하는 방법을 제안한다. 기존의 연구는 RGB 영상에서 OpenPose 등의 딥러닝 모델을 이용하여 골격 정보를 추출한 후 LSTM, GRU 등의 순환신경망 모델을 이용해 인식을 수행하였다. 제안한 방법은 카메라로부터 골격정보를 바로 전달 받아 가속도 및 거리의 2개의 시계열 특징을 추출한 후 2-Stacked Bi-LSTM 모델을 이용하여 낙상 행위를 인식하였다. 어깨, 척추, 골반 등 주요 골격을 대상으로 중심관절을 구하고 이 중심관절의 움직임 가속도와 바닥과의 거리를 특징으로 제안하였다. 추출된 특징은 Stacked LSTM, Bi-LSTM 등의 모델과 성능비교를 수행하였고 GRU, LSTM 등의 기존연구에 비해 향상된 검출 성능을 실험을 통해 증명하였다.

• 한국농공학회논문집
• /
• 제64권1호
• /
• pp.39-50
• /
• 2022

In this study, we suggested the optimal training period for predicting the streamflow using the LSTM (Long Short-Term Memory) model based on the deep learning and CMIP5 (The fifth phase of the Couple Model Intercomparison Project) future climate scenarios. To validate the model performance of LSTM, the Jinan-gun (Seongsan-ri) site was selected in this study. We comfirmed that the LSTM-based streamflow was highly comparable to the measurements during the calibration (2000 to 2002/2014 to 2015) and validation (2003 to 2005/2016 to 2017) periods. Additionally, we compared the LSTM-based streamflow to the SWAT-based output during the calibration (2000~2015) and validation (2016~2019) periods. The results supported that the LSTM model also performed well in simulating streamflow during the long-term period, although small uncertainties exist. Then the SWAT-based daily streamflow was forecasted using the CMIP5 climate scenario forcing data in 2011~2100. We tested and determined the optimal training period for the LSTM model by comparing the LSTM-/SWAT-based streamflow with various scenarios. Note that the SWAT-based streamflow values were assumed as the observation because of no measurements in future (2011~2100). Our results showed that the LSTM-based streamflow was similar to the SWAT-based streamflow when the training data over the 30 years were used. These findings indicated that training periods more than 30 years were required to obtain LSTM-based reliable streamflow forecasts using climate change scenarios.

• 대기
• /
• 제31권1호
• /
• pp.73-83
• /
• 2021

Numerical weather prediction (NWP) models play an essential role in predicting weather factors, but using them is challenging due to various factors. To overcome the difficulties of NWP models, deep learning models have been deployed in weather forecasting by several recent studies. This study adapts long short-term memory (LSTM), which demonstrates remarkable performance in time-series prediction. The combination of LSTM model input of meteorological features and activation functions have a significant impact on the performance therefore, the results from 5 combinations of input features and 4 activation functions are analyzed in 9 Automated Surface Observing System (ASOS) stations corresponding to cities/islands/mountains. The optimized LSTM model produces better performance within eight forecast hours than Local Data Assimilation and Prediction System (LDAPS) operated by Korean meteorological administration. Therefore, this study illustrates that this LSTM model can be usefully applied to very short-term weather forecasting, and further studies about CNN-LSTM model with 2-D spatial convolution neural network (CNN) coupled in LSTM are required for improvement.

• 한국융합학회논문지
• /
• 제9권12호
• /
• pp.47-52
• /
• 2018

개체명 인식(Named Entity Recognition)시스템은 문서로부터 고유한 의미를 가질 수 있는 인명(PS), 지명(LC), 기관명(OG) 등의 개체명을 추출하고 추출된 개체명의 범주를 결정하는 시스템이다. 최근 딥러닝 방식을 이용한 개체명 인식 연구에서 입력 데이터의 앞, 뒤 방향을 고려한 LSTM 기반의 Bi-LSTM 모델로부터 출력 데이터 간의 전이 확률을 이용한 CRF를 결합한 방식의 Bi-LSTM-CRF가 우수한 성능을 보이고, 문자 및 단어 단위의 효율적인 임베딩 벡터생성에 관한 연구와 CNN, LSTM을 활용한 모델에서도 좋은 성능을 보여주고 있다. 본 연구에서는 한국어 개체명 인식시스템 성능 향상을 위해 자질을 보강한 Bi-LSTM-CNN-CRF 모델에 관해 기술하고 전통문화 말뭉치구축 방식에 대해 제안한다. 그리고 구축한 말뭉치를 한국어 개체명 인식 성능 향상을 위한 자질 보강 모델 Bi-LSTM-CNN-CRF로 학습한 결과에 대해 제안한다.

• Journal of Information Processing Systems
• /
• 제15권3호
• /
• pp.694-706
• /
• 2019

In this study, we applied the long short-term memory (LSTM) model to classify the cryptocurrency price time series. We collected historic cryptocurrency price time series data and preprocessed them in order to make them clean for use as train and target data. After such preprocessing, the price time series data were systematically encoded into the three-dimensional price tensor representing the past price changes of cryptocurrencies. We also presented our LSTM model structure as well as how to use such price tensor as input data of the LSTM model. In particular, a grid search-based k-fold cross-validation technique was applied to find the most suitable LSTM model parameters. Lastly, through the comparison of the f1-score values, our study showed that the LSTM model outperforms the gradient boosting model, a general machine learning model known to have relatively good prediction performance, for the time series classification of the cryptocurrency price trend. With the LSTM model, we got a performance improvement of about 7% compared to using the GB model.