Long Short-term Memory Units Research Articles

A closed-form continuous-depth neural-based hybrid difference features re-representation network for RUL prediction

Remaining Useful Life (RUL) prediction contributes to ensuring the reliability of mechanical systems and improving their maintenance plans. Recently, prediction methods based on deep learning have undergone rapid development. However, there are significant differences between multivariate monitoring sequences and utilizing a single model for feature extraction, which leads to reaching suboptimization. Additionally, focusing solely on a specific feature dimension usually imposes notable limitations on the model. This paper proposes a Closed-form Continuous-depth neural (CfC)-based hybrid difference Features Re-representation Network (CfC-F2RN) for RUL prediction. This method comprehensively utilizes hybrid difference features through two stages: initial feature representation and feature re-representation. In the initial feature representation stage, a Long Short-Term Memory (LSTM) is employed to capture the hidden state information of each time step in sequence data. Next, a novel attention mechanism is utilized to extract the hidden states and obtain a deep feature map. In the feature re-representation stage, the encoded features are re-represented using a decoder composed of the CfC model. Finally, the hidden state of the last LSTM unit in the encoder, serving as supplementary information, is combined with the decoded features to form a dual-latent feature representation of the mixed differential features. The prediction subnetwork is then adopted to accomplish RUL prediction. The superiority of CfC-F2RN is substantiated through benchmarking against established methods using the C-MAPSS dataset.

Character-Level Text Generation for Shakespearean Style with LSTMs

This paper presents a pioneering approach to text generation employing Recurrent Neural Networks (RNN) with Long Short-Term Memory (LSTM) architecture, inspired by the rich and timeless prose of William Shakespeare. The motivation stems from the enduring allure of Shakespearean language, which has captivated audiences across centuries, and the challenge of replicating itsintricate style using modern computational techniques. Our research contributes a novel methodology that leverages the capabilities of RNN LSTM networks to emulate the linguistic nuances of Shakespeare with remarkable fidelity. The paper begins by providing a comprehensive overview of RNN LSTM networks, highlighting their suitability for sequential data processing tasks and their ability to capture long-rangedependencies. A review of related work in the field sets the stage for our proposed approach, shedding light on recent advancements and methodologies employed in text generation using similar techniques. We formulate the problem by defining the mathematical framework, optimization objectives, and evaluation metrics for our proposed model. The architecture consists of three layers: the data layer for preprocessing input text data, the intelligence layer comprising multiple LSTM units for capturing different aspects of Shakespearean language, and the application layer for generating output text based on learned representations. Experimental results demonstrate the effectiveness of our approach, with evaluations conducted on a corpus of Shakespearean texts.In conclusion, our research presents a significant advancement in the field of natural language generation, opening new avenues for exploring the intersection of literature and artificial intelligence.

Empowering Communication: A Deep Learning Framework for Arabic Sign Language Recognition with an Attention Mechanism

This article emphasises the urgent need for appropriate communication tools for communities of people who are deaf or hard-of-hearing, with a specific emphasis on Arabic Sign Language (ArSL). In this study, we use long short-term memory (LSTM) models in conjunction with MediaPipe to reduce the barriers to effective communication and social integration for deaf communities. The model design incorporates LSTM units and an attention mechanism to handle the input sequences of extracted keypoints from recorded gestures. The attention layer selectively directs its focus toward relevant segments of the input sequence, whereas the LSTM layer handles temporal relationships and encodes the sequential data. A comprehensive dataset comprised of fifty frequently used words and numbers in ArSL was collected for developing the recognition model. This dataset comprises many instances of gestures recorded by five volunteers. The results of the experiment support the effectiveness of the proposed approach, as the model achieved accuracies of more than 85% (individual volunteers) and 83% (combined data). The high level of precision emphasises the potential of artificial intelligence-powered translation software to improve effective communication for people with hearing impairments and to enable them to interact with the larger community more easily.

Research on Term Recognition in the Field of English Linguistics Based on Deep Learning Algorithms

Linguistic recognition, a fundamental aspect of natural language processing, has seen significant advancements with the integration of deep learning algorithms. his paper presents a novel approach leveraging deep learning algorithms, specifically an optimized Long Short-Term Memory (LSTM) model, for term recognition tasks in English linguistics. This paper presents a novel approach leveraging deep learning algorithms, particularly Weighted Genetic Optimization Deep Learning (WGODL), for term recognition tasks in English linguistics. his hybrid approach not only enhances model accuracy but also improves robustness and generalization, enabling accurate term recognition even in noisy or ambiguous linguistic contexts. The weighted genetic optimization mechanism strategically prioritizes relevant features during the learning process, further enhancing the model's ability to discern domain-specific terms. Additionally, qualitative analysis highlights the interpretability of learned representations, shedding light on the underlying linguistic structures captured by the model. By harnessing the sequential nature of language data, the proposed model integrates LSTM units with optimization techniques to effectively capture contextual dependencies and semantic relationships within text corpora. Through extensive experimentation on diverse linguistic datasets, our optimized LSTM model achieves superior term recognition accuracy, surpassing traditional methods by 8.3%.

Enhancing Air Quality Forecasting: A Novel Spatio-Temporal Model Integrating Graph Convolution and Multi-Head Attention Mechanism

Forecasting air quality plays a crucial role in preventing and controlling air pollution. It is particularly significant for improving preparedness for heavily polluted weather conditions and ensuring the health and safety of the population. In this study, a novel deep learning model for predicting air quality spatio-temporal variations is introduced. The model, named graph long short-term memory with multi-head attention (GLSTMMA), is designed to capture the temporal patterns and spatial relationships within multivariate time series data related to air quality. The GLSTMMA model utilizes a hybrid neural network architecture to effectively learn the complex dependencies and correlations present in the data. The extraction of spatial features related to air quality involves the utilization of a graph convolutional network (GCN) to collect air quality data based on the geographical distribution of monitoring sites. The resulting graph structure is imported into a long short-term memory (LSTM) network to establish a Graph LSTM unit, facilitating the extraction of temporal dependencies in air quality. Leveraging a Graph LSTM unit, an encoder-multiple-attention decoder framework is formulated to enable a more profound and efficient exploration of spatio-temporal correlation features within air quality time series data. The research utilizes the 2019–2021 multi-source air quality dataset of Qinghai Province for experimental assessment. The results indicate that the model effectively leverages the impact of multi-source data, resulting in optimal accuracy in predicting six air pollutants.

Modeling Of Hyperparameter Tuned RNN-LSTM and Deep Learning For Garlic Price Forecasting In Indonesia

In the Indonesian garlic industry, the unpredictability of garlic prices poses a substantial challenge, impacting the sector's stability and growth. This research aims to address this issue by developing a highly accurate predictive model using a Recurrent Neural Network (RNN) with Long Short-Term Memory (LSTM). The study employs a dataset spanning 782 days, meticulously divided with 80% dedicated to training and 20% to testing. The model, equipped with 50 LSTM units, undergoes intensive training over 100 epochs, with a batch size of 5. Its effectiveness is evaluated using the Root Mean Squared Error (RMSE) and Mean Absolute Percentage Error (MAPE), revealing exceptional predictive capabilities. The model achieves a low RMSE and MAPE in both training and testing phases, underscoring its accuracy and reliability in forecasting garlic prices. These results indicate not only the success of the RNN-LSTM model in capturing the complex patterns of price fluctuations but also highlight the potential of machine learning in enhancing time series analysis. This breakthrough offers significant implications for stakeholders in the garlic industry, providing a powerful tool for informed decision-making and strategic market planning, thereby contributing to the sector's sustainable development and stability

A Non-Invasive Load Identification Method Considering Feature Dimensionality Reduction and DB-LSTM

As the demand for detailed load data descriptions in modern power systems continues to increase, challenges such as high computational complexity in load identification tasks and high hardware requirements for devices have significantly hindered progress. Therefore, this paper proposes a non-intrusive load identification method using Densely-connected Bi-directional Long Short-Term Memory (DB-LSTM) with Kernel Principal Component Analysis. Firstly, a bilateral sliding window algorithm is employed for event detection in the data collected by load identification devices, checking for the switching on and off of electrical appliances. Secondly, after detecting the switching of load devices and extracting features, Kernel Principal Component Analysis is used to reduce data dimensions due to the complexity of existing features, selecting more relevant characteristics. Finally, a densely connected Bi-directional Long Short-Term Memory (LSTM) network is utilized. This enhances global and dynamic local features by stacking LSTM units and combining them with dense skip connections, providing additional channels for signal transmission, thereby strengthening feature propagation and reducing the number of parameters. This approach lowers computational complexity and improves the efficiency of the model’s load identification. The proposed model is compared and validated against mainstream non-intrusive load identification models through experiments, demonstrating its higher efficiency in load identification.

Application of Hybrid DeepLearning Architectures for Identification of Individuals with Obsessive Compulsive Disorder Based on EEG Data.

Objective: Obsessive-compulsive disorder (OCD) is a highly common psychiatric disorder. The symptoms of this condition overlap and co-occur with those of other psychiatric illnesses, making diagnosis difficult. The availability of biomarkers could be useful for aiding in diagnosis, although prior neuroimaging studies were unable to provide such biomarkers. Method: In this study, patients with OCD were classified from healthy controls using 2 different hybrid deep learning models: one-dimensional convolutional neural networks (1DCNN) together with long-short term memory (LSTM) and gradient recurrent units (GRU), respectively. Results: Both models exhibited exceptional classification accuracies in cross-validation and external validation phases. The mean classification accuracies in the cross-validation stage were 90.88% and 85.91% for the 1DCNN-LSTM and 1DCNN-GRU models, respectively. The inferior frontal, temporal, and occipital electrodes were predominant in providing discriminative features. Conclusion: Our findings underscore the potential of hybrid deep learning architectures utilizing EEG data to effectively differentiate patients with OCD from healthy controls. This promising approach holds implications for advancing clinical decision-making by offering valuable insights into diagnostic markers for OCD.

Improving cascade reservoir inflow forecasting and extracting insights by decomposing the physical process using a hybrid model

Accurate and reliable inflow forecasting is essential for efficient reservoir operation, which serves various purposes such as flood control, hydropower generation, water supply and irrigation. Although data-driven models have been vastly used to improve forecasting, the lack of interpretability and physical insights, particularly for neural network models, limit their trustworthiness and usability. In this study, we introduce an interpretable hybrid model to decompose inflow into precipitation-runoff and routing processes. This approach not only improves model performance but also facilitates the extraction of valuable hydrological insights. When applied to the Missouri River Basin cascade reservoir systems, the hybrid model demonstrates strong performance when compared to the state-of-the-art hydrologic model HEC-HMS and the widely used Long Short-Term Memory (LSTM) model in predicting reservoir inflow, especially for lower and smaller reservoirs in the cascade reservoir systems. For the upper and larger reservoirs, it achieves substantial improvement in Nash-Sutcliffe efficiency, ranging from 0.05 to 0.08 and 0.28 to 0.32, except for Fort Peck reservoir, when compared with the LSTM and HEC-HMS models. The decomposition of physical processes enables the hybrid model to represent each component more effectively, thereby leading to improved results in simulation and forecasting. Furthermore, the hybrid model shows how meteorological information updates the memory state of the LSTM unit and routing parameters, which in turn influences inflow dynamics, particularly in situations involving snowmelt periods. The proposed approach also allows the extraction of valuable hydrological insights, including the assessment of the contributions of precipitation-runoff processes versus routing to inflow dynamics, dynamic water travel time, and routing coefficients. The case studies have demonstrated the promising capabilities of this interpretable hybrid model in improving reservoir inflow forecasting and providing meaningful hydrological insights into reservoir inflow dynamics.

Algorithm Optimizer in GA-LSTM for Stock Price Forecasting

Fluctuating stock prices make it difficult for investors to see investment opportunities. One tool that can help investors overcome this is represented by forecasting techniques. Long Short-Term Memory (LSTM) is one of deep learning methods used in forecasting time series. The training and success of deep learning is strongly influenced by the selection of hyperparameters. This research uses a hybrid method between the Genetic Algorithm (GA) and LSTM to find a suitable model for predicting stock prices. GA is used in optimizing the architecture such as the number of epochs, window size, and the number of LSTM units in the hidden layer. Tuning optimizer is also carried out using several optimizers to achieve the best value. From method that has been applied, it shows that the method has a good level of accuracy with MAPE values below 10% in every optimizer used. The error rate generated is quite low, in case-1 with a minimum RMSE value of 93.03 and 94.40, & in case-2 with an RMSE value of 104.99 and 150.06 during training and testing. A fairly stable and small value is generated by setting it using the Adam Optimizer.

Comparing the Performance of Machine Learning and Deep Learning Algorithms in Wastewater Treatment Process

This study assessed the performance of single and modified algorithms based on machine learning and deep learning for wastewater treatment process. More specifically, this study adopted support vector machine (SVM), random forest (RF), and artificial neural network (ANN) for machine learning as well as long short-term memory (LSTM) for deep learning. The performance of these (single) algorithms were compared with that of modified ones processed through hyperparameter tuning, ensemble learning (only for machine learning), and multi-layer stacking (i.e., two layers of LSTM units). The daily effluent of wastewater treatment process observed between 2017 and 2022 in the Cheong-Ju National Industrial Complex was used as input to all tested algorithms, which was evaluated with respect to mean squared error. For the model performance evaluation, discharge and biochemical oxygen demand are selected as dependent variables out of nine measured parameters. Results showed that the performance of any machine learning algorithms was superior to their competitor LSTM. This is mainly attributed to a small amount of input data provided to the LSTM algorithm and unstable effluent wastewater characteristics. Meanwhile, hyperparameter tuning improved the performance of all tested algorithms. However, ensemble learning for machine learning and two-layer stacking for LSTM generally resulted in performance degradation as compared to that of single algorithms, regardless of dependent variables. Therefore, this calls for a careful design and evaluation of modified algorithms, specifically for model architecture and performance improvement processes.

Modified state activation functions of deep learning-based SC-FDMA channel equalization system

The most important function of the deep learning (DL) channel equalization and symbol detection systems is the ability to predict the user’s original transmitted data. Generally, the behavior and performance of the deep artificial neural networks (DANNs) rely on three main aspects: the network structure, the learning algorithms, and the activation functions (AFs) used in each node in the network. Long short-term memory (LSTM) recurrent neural networks have shown some success in channel equalization and symbol detection. The AFs used in the DANN play a significant role in how the learning algorithms converge. Our article shows how modifying the AFs used in the tanh units (block input and output) of the LSTM units can significantly boost the DL equalizer's performance. Additionally, the learning process of the DL model was optimized with the help of two distinct error-measuring functions: default (cross-entropy) and sum of squared error (SSE). The DL model's performance with different AFs is compared. This comparison is conducted using three distinct learning algorithms: Adam, RMSProp, and SGdm. The findings clearly demonstrate that the most frequently used AFs (sigmoid and hyperbolic tangent functions) do not really make a significant contribution to perfect network behaviors in channel equalization. On the other hand, there are a lot of non-common AFs that can outperform the frequently employed ones. Furthermore, the outcomes demonstrate that the recommended loss functions (SSE) exhibit superior performance in addressing the channel equalization challenge compared to the default loss functions (cross-entropy).

Hybrid deep learning model-based human action recognition in indoor environment

AbstractHuman activity recognition (HAR) is an emerging challenge among researchers. HAR has many possible uses in various fields, including healthcare, sports, and security. Furthermore, there are only a few publicly accessible datasets for classifying and recognizing physical activity in the literature, and these datasets comprise fewer activities. We created and compared our dataset with available datasets, that is, NTU-RGBD, UP-FALL, UR-Fall, WISDM, and UCI HAR. The proposed dataset consists of seven activities: eating, exercise, handshake, situps, vomiting, headache, and walking. The activities were collected from 20 people between the ages of 25 and 40 years using Kinect V2 sensor at 30 FPS. For classification, we use deep learning architectures based on convolutional neural network (CNN) and long short-term memory (LSTM). Additionally, we developed a novel hybrid deep learning model by combining a CNN, a bidirectional LSTM unit, and a fully connected layer for activity identification. The suggested model builds unique guided features using the preprocessed skeleton coordinates and their distinctive geometrical and kinematic aspects. Results from the experiment are contrasted with the performance of stand-alone CNNs, LSTMs, and ConvLSTM. The proposed model’s accuracy of 99.5% surpasses that of CNN, LSTM, and ConvLSTM, which have accuracy rates of 95.76%, 97%, and 98.89%, respectively. Our proposed technique is invariant of stance, speed, individual, clothes, etc. The proposed dataset sample is accessible to the general public.

An In-Depth Analysis of Autism Spectrum Disorder Using Optimized Deep Recurrent Neural Network

Autism spectrum disease is one of the severe neuro developmental disorders that are currently present worldwide (ASD). It is a chronic disorder that has an impact on a person’s behaviour and communication abilities. The world health organization’s 2019 study states that an increasing number of people are being diagnosed with ASD, which poses a risk because it is comparable to high medical expenses. Early detection can significantly lessen the impact. Traditional techniques are costly and time-consuming. This paper offers a Novel Deep Recurrent Neural Network (NDRNN) algorithm for the detection of the level of autism to address the aforementioned problems. The deep recurrent neural network is developed with several hidden recurrent network layers with Long-Short Term Memory (LSTM) units. In this work, Artificial Algae Algorithm (AAA) is used as a feature extraction algorithm, to obtain the best optimal features among the listed feature set. An Intelligent Water Droplet (IWD) algorithm is used for obtaining optimal weight and bias value for the recurrent neural network. The algorithm was evaluated for the dataset obtained by the Indian scale for assessment of autism. Experimental results shows that this proposed model produces the 91% of classification accuracy and 92% of sensitivity and reduces the cost.

An Intrusion Detection Method based on Fusion Neural Network

Aiming at the problems of class imbalance, insufficient feature learning, weak generalization ability, and representation capability in existing intrusion detection models, we propose a multi-scale feature fusion Intrusion Detection Model (MSFF). This model combines multi-scale one-dimensional convolution and bidirectional long short-term memory (LSTM) networks, and incorporates residual connections with identity mappings to address the problem of network degradation. The multi-scale convolution captures feature representations at different levels, thereby improving the expressive power of the model. The WGAN-GP algorithm is employed to augment the minority samples and balance the dataset. By performing convolution operations and extracting local window features and global features using bidirectional LSTM units, the model effectively captures temporal information and long-term dependencies. Experimental results demonstrate significant performance improvement compared to a single model. The MSFF model achieves an accuracy of 99.50% and 94.73% in binary classification experiments on the NSL-KDD and UNSW-NB15 datasets, respectively, and an accuracy of 99.50% and 83.78% in multi-class classification experiments.

Accelerating AI-Based Battery Management System’s SOC and SOH on FPGA

Lithium battery-based electric vehicles (EVs) are gaining global popularity as an alternative to combat the adverse environmental impacts caused by the utilization of fossil fuels. State of charge (SOC) and state of health (SOH) are vital parameters that assess the battery’s remaining charge and overall health. Precise monitoring of SOC and SOH is critical for effectively operating the battery management system (BMS) in a lithium battery. This article presents an experimental study for the artificial intelligence (AI)-based data-driven prediction of lithium battery parameters SOC and SOH with the help of deep learning algorithms such as Long Short-Term Memory (LSTM) and bidirectional LSTM (BiLSTM). We utilized various gradient descent optimization algorithms with adaptive and constant learning rates with other default parameters. Compared between various gradient descent algorithms, the selection of the optimal one depends on mean absolute error (MAE) and root mean squared error (RMSE) accuracy. We developed an LSTM and BiLSTM model with four hidden layers with 128 LSTM or BiLSTM units per hidden layer that use Panasonic 18650PF Li-ion dataset released by NASA to predict SOC and SOH. Our experimental results advise that the selection of the optimal gradient descent algorithm impacts the model’s accuracy. The article also addresses the problem of overfitting in the LSTM/BiLSTM model. BiLSTM is the best choice to improve the model’s performance but increase the cost. We trained the model with various combinations of parameters and tabulated the accuracies in terms of MAE and RMSE. This optimal LSTM model can predict the SOC of the lithium battery with MAE more minor than 0.0179%, RMSE 0.0227% in the training phase, MAE smaller than 0.695%, and RMSE 0.947% in the testing phase over a 25°C dataset. The BiLSTM can predict the SOC of the 18650PF lithium battery cell with MAE smaller than 0.012% for training and 0.016% for testing. Similarly, using the Adam optimization algorithm, RMSE for training and testing is 0.326% and 0.454% over a 25°C dataset, respectively. BiLSTM with an adaptive learning rate can improve performance. To provide an alternative solution to high power consuming processors such as central processing unit (CPU) and graphics processing unit (GPU), we implemented the model on field programmable gate Aarray (FPGA) PYNQ Z2 hardware device. The LSTM model using FPGA performs better.

Spatial–temporal features-based EEG emotion recognition using graph convolution network and long short-term memory

Objective. Emotion recognition on the basis of electroencephalography (EEG) signals has received a significant amount of attention in the areas of cognitive science and human–computer interaction (HCI). However, most existing studies either focus on one-dimensional EEG data, ignoring the relationship between channels, or only extract time–frequency features while not involving spatial features. Approach. We develop spatial–temporal features-based EEG emotion recognition using a graph convolution network (GCN) and long short-term memory (LSTM), named ERGL. First, the one-dimensional EEG vector is converted into a two-dimensional mesh matrix, so that the matrix configuration corresponds to the distribution of brain regions at EEG electrode locations, thus to represent the spatial correlation between multiple adjacent channels in a better way. Second, the GCN and LSTM are employed together to extract spatial–temporal features; the GCN is used to extract spatial features, while LSTM units are applied to extract temporal features. Finally, a softmax layer is applied to emotion classification. Main results. Extensive experiments are conducted on the A Dataset for Emotion Analysis using Physiological Signals (DEAP) and the SJTU Emotion EEG Dataset (SEED). The classification results of accuracy, precision, and F-score for valence and arousal dimensions on DEAP achieved 90.67% and 90.33%, 92.38% and 91.72%, and 91.34% and 90.86%, respectively. The accuracy, precision, and F-score of positive, neutral, and negative classifications reached 94.92%, 95.34%, and 94.17%, respectively, on the SEED dataset. Significance. The above results demonstrate that the proposed ERGL method is encouraging in comparison to state-of-the-art recognition research.

An accurate traffic flow prediction using long-short term memory and gated recurrent unit networks

Congestion on roadways is an issue in many cities, especially at peak times, which causes air and noise pollution and cause pressure on citizens. So, the implementation of intelligent transportation systems (ITSs) is a very important part of smart cities. As a result, the importance of making accurate short-term predictions of traffic flow has significantly increased in recent years. However, the current methods for predicting short-term traffic flow are incapable of effectively capturing the complex non-linearity of traffic flow that affects prediction accuracy. To overcome this problem, this study introduces two novel models. The first model uses two long-short term memory (LSTM) units that can extract the traffic flow temporal features followed by four dense layers to perform the traffic flow prediction. The second model uses two gated recurrent unit (GRU) units that can extract the traffic flow temporal features followed by three dense layers to perform the traffic flow prediction. The two proposed models give promising results on performance measurement system (PEMS), traffic and congestions (TRANCOS) dataset that is firstly used as metadata. So, the two models can do this in specific cases and are able to suddenly capture trend changes.

Real-time determination of moisture content of potato slices during drying using long short-term memory from image morphometric parameters

This article uses digital image processing to investigate the change in morphometric attributes of potato slices of 1 mm thickness for different drying temperatures (45, 50, 55 and 60 °C). For this purpose, a compact hot air dryer has been developed, having the desired provision for embedding an image acquisition system. The long short-term memory (LSTM) technique has been employed to determine the product moisture content inside the hot air dryer purely on the basis of morphometric characteristics of individual slices. The best-performing network optimized after hyperparameter tuning has an architecture of 156 hidden neurons, 2 LSTM units, and a learning rate of 0.0160. The optimized network’s training coefficient of determination (R2) and root mean square error (RMSE) are 0.977 and 0.0461, respectively. In addition, the quantitative uncertainty analysis using the quantile score (Q) and the prediction-interval-normalized root-mean-square width (PINRW) was carried out to evaluate the robustness of the network. This approach is instrumental in designing a non-invasive quality control system during drying phenomena. It also helps in studying the complexity of the drying mechanism by visualizing the morphometric changes taking place in the product during drying.

PREDIKSI TINGKAT TEMPERATUR KOTA SEMARANG MENGGUNAKAN METODE LONG SHORT-TERM MEMORY (LSTM)

Temperature is one of the most important attributes of climate, temperature affects life in many different ways such as in agriculture, aviation, energy, and life in general. Temperatur prediction is needed to make the right step to prevent the negative impact of climate change. Long Short-Term Memory (LSTM) is the method that can predict time series data, using unique design of neural network, LSTM can help to prevent vanishing gradient from happening which allows us to use more data from the past to predict the future. Hyperparameters like LSTM unit, epochs, and batch size are used to make the best model, the best model is the one with the lowest loss function. This research used climate data from 1 January 2019 until 31 December 2021 consist of 1096 data in total. The best prediction in this research is made by the model with 1 hidden layer, 64 LSTM units, 8 batch sizes, and 400 epochs with the lowest loss function 0.012769, this model give MAPE value of 1.97325 % and RMSE value of 0.683.