A Hybrid CNN-LSTM Model With Word-Emoji Embedding For Improving The Twitter Sentiment Analysis on Indonesia's PPKM Policy

Pane, Syafrial Fachri; Ramdan, Jenly; Putrada, Aji Gautama; Fauzan, Mohamad Nurkamal; Awangga, Rolly Maulana; Alamsyah, Nur

doi:10.1109/icitisee57756.2022.10057720

Cited by 8 publications

(3 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Then, the final result of the model is a majority vote from all weak learners [19]. Finally, CNN is a type of deep learning where one type of sublayer is the convolutional layer, namely a kernel that carries out the convolution process on the input data [20]. Here, we use 1D-CNN because our data is in the form of a time series signal.…”

Section: The Cima Windowing Algorithmmentioning

confidence: 99%

Human Activity Recognition Improvement on Smartphone Accelerometers Using Cima

Putrada,

Abdurohman,

Perdana

et al. 2024

tektrika

View full text Add to dashboard Cite

Human activity recognition (HAR) is a research field that focuses on detecting user activities and has wide applications. However, the problems that need to be solved are real-time constraints and imbalanced datasets due to different activity frequencies. Our research aims to apply classification integrated moving averages (CIMA) to HAR by evaluating its performance regarding real-time constraints and imbalanced datasets. We achieved the smartphone accelerometer dataset from Kaggle, which consists of several activities: walking, jogging, climbing, and descending stairs. We develop a general CIMA windowing algorithm with hyperparameters J and W. We benchmark CIMA with two state-of-the-art HAR methods: distributed online activity recognition system (DOLARS) and convolutional neural network (CNN). We conducted some imbalance and model size analysis. The test results show that, with J = 10 and W = 240, CIMA performs better than DOLARS and CIMA with recall, precision, and f1-score of 0.996, 0.993, and 0.994. We also prove that CIMA, assisted by quantization, has the smallest model size compared to the CNN and DOLARS model sizes. Finally, we demonstrate that CIMA performs well for imbalanced datasets, where CIMA’s recall on upstairs and downstairs activities is better than DOLARS and CNN, with values of 1.00 and 0.98, respectively. Key Words: classification integrated moving average, human activity recognition, smartphone, accelerometer, imbalanced dataset

show abstract

Section: The Cima Windowing Algorithmmentioning

confidence: 99%

Human Activity Recognition Improvement on Smartphone Accelerometers Using Cima

Putrada,

Abdurohman,

Perdana

et al. 2024

tektrika

View full text Add to dashboard Cite

show abstract

“…If the activation function to be tested is PReLU, we use the PReLU module of Keras to activate it. We present the tanh activation function in formula (2), ReLU in formula (3), Sigmoid in formula (4) and PReLU in formula (5).…”

Section: E Lstm Model Buildingmentioning

confidence: 99%

“…By investigating the potential use of the Long Short-Term Memory (LSTM) method with Parametric Rectified Linear Unit (PReLU). activation function, this research seeks to make a significant contribution in improving the accuracy and reliability of traffic density prediction amidst the increasing vehicle growth [5].…”

Section: Introductionmentioning

confidence: 99%

IMPROVING TRAFFIC DENSITY PREDICTION USING LSTM WITH PARAMETRIC ReLU (PReLU) ACTIVATION

Alamsyah,

Yoga,

Budiman

2024

jitk

View full text Add to dashboard Cite

In the presence of complex traffic flow patterns, this research responds to the challenge by proposing the application of the Long Short-Term Memory (LSTM) model and comparing four different activation functions, namely tanh, ReLU, sigmoid, and PReLU. This research aims to improve the accuracy of traffic flow prediction through LSTM model by finding the best activation function among tanh, relu, sigmoid, and PReLU. The method used starts from the collection of traffic flow datasets covering the period 2015-2017 used to train and evaluate the LSTM model with the four activation functions. Tests were conducted by observing the Train Mean Squared Error (MSE) and Validation MSE. The experimental results show that PReLU provides the best results with a Train MSE of 0.000505 and Validation MSE of 0.000755. Although tanh, ReLU, and sigmoid provided competitive results, PReLU stood out as the optimal choice to improve the adaptability of the model to complex traffic flow patterns.

show abstract