Klasifikasi Aktivitas Manusia Menggunakan Metode Long Short-Term Memory

Penulis

  • Latansa Nurry Izza Afida Universitas Brawijaya, Malang
  • Fitra Abdurrachman Bachtiar Universitas Brawijaya, Malang
  • Imam Cholissodin Universitas Brawijaya, Malang

DOI:

https://doi.org/10.25126/jtiik.20241127060

Abstrak

Klasifikasi aktivitas manusia merupakan salah satu topik penelitian yang penting karena dapat diterapkan pada berbagai bidang dan memiliki manfaat yang luas. Penelitian mengenai klasifikasi aktivitas manusia sebelumnya telah banyak dikembangkan dengan menerapkan dataset publik pada repositori dataset Human Activity Recognition. Namun dataset tersebut memiliki fitur yang berdimensi tinggi sehingga dataset memiliki dimensi yang tinggi pula. Pada beberapa penelitian sebelumnya menunjukkan bahwa algoritma SVM dan Random Forest merupakan algoritma dengan nilai akurasi yang lebih unggul dibandingkan dengan model lainnya. Akan tetapi berdasarkan penelitian tersebut model tersebut belum pernah diimplementasikan pada kasus riil yaitu pada perangkat bergerak. Penelitian ini mengusulkan model pengenalan aktivitas manusia dengan kasus riil dengan dataset primer yang dikumpulkan dengan menggunakan smartphone. Pengambilan dataset primer melibatkan 10 responden. Data yang terkumpul dengan smartphone direkam melalui sensor menghasilkan dataset berbentuk data time series. Dataset primer yang digunakan masih memiliki nilai yang besar dan kurangnya keseimbangan jumlah label kelas sehingga eksperimen dimulai dengan tahapan preprocessing yang dilakukan dengan menggunakan moving average untuk mereduksi data tanpa menghilangkan informasi. Selain itu juga dilakukan SMOTE untuk menyeimbangkan jumlah masing - masing kelas data. Data latih memiliki proporsi sebanyak 80%, data validasi sebanyak 10% dan data uji sebanyak 10%. Penelitian ini menggunakan LSTM untuk klasifikasi aktivitas manusia karena algoritma ini sangat baik untuk memproses data time series berjumlah banyak. Hasil klasifikasi kemudian dibandingkan dengan algoritma terbaik pada beberapa penelitian sebelumnya. Hasil eksperimen didapatkan bahwa model LSTM dapat mengungguli model SVM dan Random Forest. Hasil klasifikasi menggunakan algoritma LSTM mencapai akurasi, Precision, Recall, dan F1-score 95%, 96%, 95%, dan 95%, secara berurutan.

 

Abstract

Human activity classification is one of the important research topics because it can be applied to various fields and have broad benefits. Research on human activity classification has previously been developed by applying public datasets to the available Human Activity Recognition dataset repository. However, the dataset has high dimensional features so that the dataset has high dimensions as well. Previous study has shown that SVM and Random Forest algorithms are algorithms with superior accuracy values compared to other models. However, based on previous research, the model has never been implemented in real cases, namely on mobile devices. This research proposes a human activity recognition model in real cases situation with primary datasets collected using smartphones. The data collection for the dataset involved 10 respondents. The data collected using a smartphone recorded via sensors to produce a dataset in the form of time series data. The primary dataset used still has a large value and there is a lack of balance in the number of class labels. To this end, the experiment begins with a preprocessing stage which is carried out using a moving average to reduce the data without losing information. In addition, SMOTE was also carried out to balance the number of each data class. The proportion of training data, validation data, and testing data is 80%, 10%, and 10%, respectively. This research uses LSTM for human activity classification because this algorithm is very good for processing large amounts of time series data. The classification results were then compared with the best algorithms in several previous studies. Experimental results show that the LSTM model can outperform the SVM and Random Forest models. Classification results using the LSTM algorithm reached Accuracy, Precision, Recall, dan F1-score 95%, 96%, 95%, and 95%, respectively.

Downloads

Download data is not yet available.

Referensi

ANGUITA D., GHIO. A., & ONETO L, 2013. A Public Domain Dataset for Human Activity Recognition Using Smartphones. European Sympsosium on Artificial Neural Network, Computational Intelligent and Machine Learning.

M. ULLAH, H. ULLAH, S. D. KHAN and F. A. CHEIKH 2019. Stacked Lstm Network for Human Activity Recognition Using Smartphone Data. 8th European Workshop on Visual Information Processing (EUVIP), Roma, Italy, 2019, pp. 175-180, doi: 10.1109/EUVIP47703.2019.8946180.

R. MUTAGEKI and D. S. HAN. 2020.A CNN-LSTM Approach to Human Activity Recognition. International Conference on Artificial Intelligence in Information and Communication (ICAIIC), Fukuoka, Japan, 2020, pp. 362-366,

doi: 10.1109/ICAIIC48513.2020.9065078.

S. YU and L. QIN. 2018. Human Activity Recognition with Smartphone Inertial Sensors Using Bidir-LSTM Networks. 3rd International Conference on Mechanical, Control and Computer Engineering (ICMCCE), Huhhot, China, 2018, pp. 219-224.

doi: 10.1109/ICMCCE.2018.00052.

SAGHEER, A., & KOTB, M. 2019. Time series forecasting of petroleum production using deep LSTM recurrent networks. Neurocomputing, 323, 203–213.

doi: 10.1016/j.neucom.2018.09.082

CHOLISSODIN, I. K. PINASTHIKA., J. H. PRASTIWI., & R. DEWANTARA., 2021. Social Computing to Create Goverment Public Policy Document Blueprint Draft Based on Social Media Data About COVID-19 Using LSTM and MMR Hybrid Algorithm. International Conference on Green Technology, p. 2.

YU, S., & QIN, L. (2018). Human activity recognition with smartphone inertial sensors using bidir-LSTM networks. Proceedings - 2018 3rd International Conference on Mechanical, Control and Computer Engineering, ICMCCE 2018, 219–224.

KHAN, A. M., LEE, Y. K., LEE, S. Y., & KIM, T. S. 2010. A triaxial accelerometer-based physical-activity recognition via augmented-signal features and a hierarchical recognizer. IEEE Transactions on Information Technology in Biomedicine, 14(5), 1166–1172. https://doi.org/10.1109/TITB.2010.2051955

STITSON, M. O., WESTON, J. A. E., GAMMERMAN, A., VOVK, V., & VAPNIK, V. 1996. Theory of Support Vector Machines: Vol. +.

RABBI, J., FUAD, MD. T. H., & AWAL, MD. A. (2021). Human Activity Analysis and Recognition from Smartphones using Machine Learning Techniques. http://arxiv.org/abs/2103.16490

BEVILACQUA, A., MACDONALD, K., RANGAREJ, A., WIDJAYA, V., CAULFIELD, B., & KECHADI, T. 2019. Human Activity Recognition with Convolutional Neural Netowrks. https://doi.org/10.1007/978-3-030-10997-4_33

SUTOYO, E., ASRI FADLURRAHMAN, M., Telekomunikasi Jl Terusan Buah Batu, J., Dayeuhkolot, K., Bandung, K., & Barat, J. (n.d.). JEPIN (Jurnal Edukasi dan Penelitian Informatika) Penerapan SMOTE untuk Mengatasi Imbalance Class dalam Klasifikasi Television Advertisement Performance Rating Menggunakan Artificial Neural Network.

CHAWLA, N. V, BOWYER, K. W., HALL, L. O., & KEGELMEYER, W. P. 2002. SMOTE: Synthetic Minority Over-sampling Technique. In Journal of Artificial Intelligence Research (Vol. 16).

KHALILZADEH, J., & TASCI, A. D. A. 2017. Large sample size, significance level, and the effect size: Solutions to perils of using big data for academic research. Tourism Management, 62, 89–96. https://doi.org/10.1016/j.tourman.2017.03.026

REISS, A., HENDEBY, G., & STRICKER, D. 2013. A Competitive Approach for Human Activity Recognition on Smartphones (Vol. 2013). http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-97302

SAGHEER, A., & KOTB, M. 2019. Time series forecasting of petroleum production using deep LSTM recurrent networks. Neurocomputing, 323, 203–213. https://doi.org/10.1016/j.neucom.2018.09.082

AINI, N., SINURAT, S., & ADELINA HUTABARAT, S. 2018. Penerapan Metode Simple Moving Average Untuk Memprediksi Hasil Laba Laundry Karpet Pada CV. Homecare. In Jurnal Riset Komputer (JURIKOM) (Vol. 5, Issue 2). http://ejurnal.stmik-budidarma.ac.id/index.php/jurikom|Page|167

ZHANG, M., & SAWCHUK, A. A. 2012. A feature selection-based framework for human activity recognition using wearable multimodal sensors. BODYNETS 2011 - 6th International ICST Conference on Body Area Networks, 92–98. https://doi.org/10.4108/icst.bodynets.2011.247018

YULIANTI, R., PASEK, G., WIJAYA, S., & BIMANTORO, D. F. (N.D.). Pengenalan Pola Tulisan Tangan Suku Kata Aksara Sasak Menggunakan Metode Moment Invariant dan Support Vector Machine (Handwritten Sasak Ancient Script Recognition using Moment Invariant and Support Vector Machine). http://jcosine.if.unram.ac.id/

JISHAN, S. T., RASHU, R. I., HAQUE, N., & RAHMAN, R. M. 2015. Improving accuracy of students’ final grade prediction model using optimal equal width binning and synthetic minority over-sampling technique. Decision Analytics, 2(1). https://doi.org/10.1186/s40165-014-0010-2

GHATE, V., & HEMALATHA C, S. 2023. A comprehensive comparison of machine learning approaches with hyper-parameter tuning for smartphone sensor-based human activity recognition. Measurement: Sensors, 30. https://doi.org/10.1016/j.measen.2023.100925

BACHTIAR, F. A., ARIFIEN, Z., PUTRA, A. R., & AKBAR, P. I. 2021. Feature Selection with Particle Swarm Optimization for Human Activity Recognition Using Learning Vector Quantization. ACM International Conference Proceeding Series, 132–138. https://doi.org/10.1145/3479645.3479693

Unduhan

Diterbitkan

26-08-2024

Terbitan

Bagian

Ilmu Komputer

Cara Mengutip

Klasifikasi Aktivitas Manusia Menggunakan Metode Long Short-Term Memory. (2024). Jurnal Teknologi Informasi Dan Ilmu Komputer, 11(2), 357-368. https://doi.org/10.25126/jtiik.20241127060