Using Artificial Neural Networks in Total Manganese Estimation for Some Soils in Middle and Northern of Iraq | ||
Journal of Tikrit University For Agriculture Sciences | ||
Article 1, Volume 16, Issue 2, June 2016, Pages 185-196 | ||
Author | ||
Mohammed Tahir Said Khalil | ||
Abstract | ||
The study was applied on 111 soil samples were collected from 22 selected locations in middle and northern Iraq differ in their total manganese contents. Aim is to use artificial neural networks to find the most efficient mathematical model for total manganese estimation. Some soil characteristics (EC, organic matter, CEC, clay content, CaCO3, pH) have been depended as inputs for assumed artificial neural networks model. Statistically, stepwise multiple linear regression model was carried out to find a linear harmony between artificial neural network outputs and observed total manganese data for soil samples were excluded from training group. Results of neural networks application referred that root mean square error (RMSE= 50.3), mean absolute percent error (MAPE= 8.6%), validation factor (VF= 90.5) and correlation relationship (r= 0.92) between observed and estimated soil total manganese values, also statistical analysis by using MATLAB programs band for neural networks outputs referred that correlation relationships for training, validation, test groups and total data of soil samples were (r= 0.88) , (r= 0.92) , (r= 0.85) , (r= 0.88) respectively indicating for efficiency artificial neural networks in total manganese estimation with depending upon soil characteristics mentioned above. | ||
Keywords | ||
Total Mn; ANN; Tan; sigmoid function; linear function; Training; validation; test groups | ||
Statistics Article View: 116 PDF Download: 66 |