Application of hybrid algorithm of bionic heuristic and machine learning in nonlinear sequence

Qian, Lipeng; Liu, Changzheng; Yi, Jiaxin; Liu, Shannan

doi:10.1088/1742-6596/1682/1/012009

Cited by 5 publications

(6 citation statements)

References 38 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…From Figure 1, Figure 2 (b), and Eq. ( 6), the outer gradient of Layer 3 stated by: (7) The inner gradient of Layer 3 stated by: (8) The local error of Layer 3 stated by: (9) The outer gradient of Layer 2 stated by: (10) The inner gradient of Layer 2 stated by: (11) The local error of Layer 2 stated by: (12) The outer gradient of Layer 1 stated by: (13) The inner gradient of Layer 1 stated by: (14) The local error of Layer 1 stated by: (15) The constants of are the biases of each layer, while is the local gradient stabilizer constant with adjustable value in the range . The network weight adjustment for each process iteration uses the Delta Rule algorithm [18] which is stated by: (16) is the learning rate with adjustable value in the range .…”

Section: B Heuristic Networkmentioning

confidence: 99%

“…It can predict the distribution of the percentage of GDP (Gross Domestic Product) in [11], whose problem-solving method has been changed using DNN (Deep Neural Network) in [12]. This heuristic network is very different from the heuristic approach used, such as in [9,10,[13][14][15][16]. The Heuristic Network is almost similar to ANN (Artificial Neural Network) but differs in the learning process.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A Heuristic Network for Electricity Demand Projection

Mulyanto¹,

Suprapty²,

Malani³

et al. 2022

Proceedings of the International Conference on Applied Science and Technology on Social Science 2022 (iCAST-SS 2022)

View full text Add to dashboard Cite

Indonesia's electricity demand in 2025 has been projected by the 2016 Indonesia Energy Outlook (OEI) of 513 TWh based on data obtained during the 2000-2015 period. This study tries to project the total electricity demand during 2016-2025 based on data on nominal GRDP growth, population, electricity sales in the household, commercial and industrial sectors, and electrification ratio. All data obtained in the form of time series data were modeled using MISO-ARX (Multi-Input Single Output Auto Regressive with Exogenous Input). This model is then represented using a Heuristic Network. After the training process, this network is used to predict the growth of total electricity sales in 2016-2025 based on growth data from 2001 -2015. The prediction results are used to project electricity demand in 2016-2025. The comparison of the projection results using the Heuristic Network with the results of the OEI-2016 project is MAPE = 4.46%. With this relatively small MAPE, the projection results using the Heuristic Network can be considered fairly close to the results of the OEI-2016 projection

show abstract

Section: B Heuristic Networkmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

A Heuristic Network for Electricity Demand Projection

Mulyanto¹,

Suprapty²,

Malani³

et al. 2022

Proceedings of the International Conference on Applied Science and Technology on Social Science 2022 (iCAST-SS 2022)

View full text Add to dashboard Cite

show abstract

“…Qian et al . 17 stated that standalone ML models are also less capable of coupling and processing nonlinear problems compared to hybrid ML models. Several other disadvantages, depending on the standalone ML algorithm used, are overfitting, lack of memory, parameter uncertainty, cognitive uncertainties, and local minimization drawback, the requirement to comply with data assumptions, ability to only provide linear solutions, assumption of independence between features, and requirement of large data samples to achieve good performance 15 , 18 .…”

Section: Literature Reviewmentioning

confidence: 99%

“…Qian et al . 17 stated that the training time of hybrid ML models is high, especially when dealing with complex problems. Hybrid ML models require many more input parameters to be considered during training compared to standalone ML models.…”

Section: Literature Reviewmentioning

confidence: 99%

“…Hybrid ML models require many more input parameters to be considered during training compared to standalone ML models. This often restricts the development and usage of hybrid ML models 17 . In addition, complicated architecture and an unknown optimal number of clusters have also been reported as disadvantages of utilizing hybrid ML models 15 .…”

Section: Literature Reviewmentioning

confidence: 99%

See 1 more Smart Citation

Suspended sediment load prediction using long short-term memory neural network

AlDahoul

Essam

Kumar

et al. 2021

Sci Rep

View full text Add to dashboard Cite

Rivers carry suspended sediments along with their flow. These sediments deposit at different places depending on the discharge and course of the river. However, the deposition of these sediments impacts environmental health, agricultural activities, and portable water sources. Deposition of suspended sediments reduces the flow area, thus affecting the movement of aquatic lives and ultimately leading to the change of river course. Thus, the data of suspended sediments and their variation is crucial information for various authorities. Various authorities require the forecasted data of suspended sediments in the river to operate various hydraulic structures properly. Usually, the prediction of suspended sediment concentration (SSC) is challenging due to various factors, including site-related data, site-related modelling, lack of multiple observed factors used for prediction, and pattern complexity.Therefore, to address previous problems, this study proposes a Long Short Term Memory model to predict suspended sediments in Malaysia's Johor River utilizing only one observed factor, including discharge data. The data was collected for the period of 1988–1998. Four different models were tested, in this study, for the prediction of suspended sediments, which are: ElasticNet Linear Regression (L.R.), Multi-Layer Perceptron (MLP) neural network, Extreme Gradient Boosting, and Long Short-Term Memory. Predictions were analysed based on four different scenarios such as daily, weekly, 10-daily, and monthly. Performance evaluation stated that Long Short-Term Memory outperformed other models with the regression values of 92.01%, 96.56%, 96.71%, and 99.45% daily, weekly, 10-days, and monthly scenarios, respectively.

show abstract