Analytical Study of Time Series Generation by Feed-Forward Networks

Multilayer feed-forward neural networks are widely used based on minimization of an error function. Back propagation (BP) is a famous training method used in the multilayer networks but it often suffers from the drawback of slow convergence. To make the learning faster, we propose 'Fusion of Activation Functions' (FAF) in which different conventional activation functions (AFs) are combined to compute final activation. This has not been studied extensively yet. One of the sub goals of the paper is to check the role of linear AFs in combination. We investigate whether FAF can enable the learning to be faster. Validity of the proposed method is examined by performing simulations on challenging nine real benchmark classification and time series prediction problems. The FAF has been applied to 2-bit, 3-bit and 4-bit parity, the breast cancer, Diabetes, Heart disease, Iris, wine, Glass and Soybean classification problems. The algorithm is also tested with Mackey-Glass chaotic time series prediction problem. The algorithm is shown to work better than other AFs used independently in BP such as sigmoid (SIG), arctangent (ATAN), logarithmic (LOG).

Download Full-text

Modeling of monthly rainfall and runoff of Urmia lake basin using “feed-forward neural network” and “time series analysis” model

Water Resources and Industry ◽

10.1016/j.wri.2014.10.003 ◽

2014 ◽

Vol 7-8 ◽

pp. 38-48 ◽

Cited By ~ 44

Author(s):

Jamileh Farajzadeh ◽

Ahmad Fakheri Fard ◽

Saeed Lotfi

Keyword(s):

Neural Network ◽

Time Series ◽

Time Series Analysis ◽

Monthly Rainfall ◽

Urmia Lake ◽

Lake Basin ◽

Analysis Model ◽

Feed Forward Neural Network ◽

Feed Forward ◽

Urmia Lake Basin

Download Full-text

PERAMALAN INFLOW UANG KARTAL BANK INDONESIA KPW TASIKMALAYA JAWA BARAT DENGAN METODE KLASIK DAN MODERN

Jurnal Statistika Universitas Muhammadiyah Semarang ◽

10.26714/jsunimus.8.2.2020.166-174 ◽

2020 ◽

Vol 8 (2) ◽

pp. 166

Author(s):

Nur Silviyah Rahmi

Keyword(s):

Neural Network ◽

Time Series ◽

Multilayer Perceptron ◽

Feed Forward Neural Network ◽

Feed Forward

Ketersediaan uang kartal di Bank Indonesia (BI) dapat ditinjau melalui arus keluar masuknya uang kartal yang disebut dengan istilah inflow. Banyaknya uang yang beredar di masyarakat akan berpengaruh pada kondisi perekonomian suatu negara, sehingga Bank Indonesia (BI) menyusun perencanaan kebutuhan uang rupiah. Penelitian ini bertujuan untuk meramalkan inflow uang kartal di KPw Bank Indonesia (BI) Tasikmalaya dengan menggunakan pemodelan ARIMA, ARIMAX, Metode Dekomposisi, Metode Winter’s, MLP (Multilayer Perceptron) atau FFNN (Feed Forward Neural Network), Regresi Time Series, Metode Naïve dan Model Hybrid. Dari delapan metode runtun waktu tersebut baik klasik maupun modern akan dicari metode mana yang memberikan hasil akurasi ramalan yang terbaik dengan kriteria RMSE, MAPE dan MAD. Kesimpulan yang dihasilkan yaitu Hybrid ARIMA-NN yang merupakan gabungan dari model ARIMA dengan neural network tidak menjamin kinerja hasil peramalan yang lebih baik. Seperti yang disebutkan dalam hasil M3 Competition, semakin kompleks metode yang digunakan belum tentu metode tersebut menghasilkan akurasi yang lebih baik dibandingkan metode sederhana (klasik). Pada ramalan data inflow KPw BI Tasikmalaya Jawa Barat ini, menghasilkan kesimpulan bahwa metode regresi time series memiliki nilai kriteria pemodelan paling kecil dibandingkan dengan metode lainnya.

Download Full-text