Pengaruh Variasi Jumlah Neuron dalam Hidden Layer Algoritma Pelatihan Levenberg-Marquardt Jaringan Backpropagation: A Systematic Literature Review

<p><strong>This analysis is done to determine and is a consideration for future research related to different types of problem solving by using the training algorithm of Backpropagation network. This study uses 4 steps selections in filter articles that will be used in literary studies, namely 1) Identification 2) Screening 3) Eligibility and 4) Included. The number of items filtered in this study is 73 articles. The article was filtered through the identification phase with a total of 205 articles, then in the screening process by assimilating the title and summary, then the eligible process with many articles filtered by 132 articles did not meet the requirements to get the final results of 73 articles for the analysis process. The number of nerve cells indicates that there is no rules that are determined related to the exact quantity of nerve cells in the hidden layer depending on all research needs and parameters applied in research. Although in some articles, the accuracy value is not briefly mentioned that the Levenberg-Marquardt training algorithm is effective in solving problems, in 21 articles filtered that the Levenberg- Marquardt training algorithm </strong><strong>has an accuracy rate of over 90%, indicating that this algorithm can be an alternative choice as a problem-solving tool due to its effectiveness and optimal accuracy of results</strong>.<strong></strong></p><p><strong><em>Keywords</em></strong> - <em>Accuracy, Backpropagation, , Effectiveness, Hidden Layer, </em>Levenberg-Marquardt<em> Algorithm</em></p>