查詢結果分析
來源資料
頁籤選單縮合
| 題 名 | 應用類神經網路與線性回歸模式預測乾稻穀淨重之研究=Application of Neural Network and Linear Regression Models on Prediction of the Net Weight of Rice Grain |
|---|---|
| 作 者 | 林聖泉; 何榮祥; 洪傳儀; | 書刊名 | 農林學報 |
| 卷 期 | 46:4 1997.12[民86.12] |
| 頁 次 | 頁49-58 |
| 分類號 | 434.1 |
| 關鍵詞 | 稻穀; 類神經網路模式; 線性回歸模式; Rice grain; Neural network model; Linear regression model; |
| 語 文 | 中文(Chinese) |
| 中文摘要 | 本研究利用前饋式類神經網路與線性回歸等兩種模式預測濕穀乾燥後之稻穀淨重 ，並比較兩者之特點。前饋式類神經網路模型由輸入層、隱藏層、與輸出層等神經層所構成 ，其中 輸入層有2 估神經元(輸入數據為濕穀含水率與容積 重)，隱藏層有1個神經元，輸 出層則為1個神經元(輸出數據為乾穀淨重)，利用倒傳遞演算法訓練網路，使預測總誤差收 斂至可接受之範圍。線性回歸模式，則分別以一 次與二次方程式表示。本研究所建立乾穀 淨重預測模式，預測稻穀淨重與實際量測結果比較，不論類神經網路模式或線性回歸模式， 所得之平均相對誤差均在 2.0% 以下；採用線性致動函數之類神經網路模式，可視為線性回 歸一次模式。本研究所發展之模式同時考慮濕穀含水率與容積重，所得之乾穀淨重預測值， 較農會採用只考慮濕穀含水率之換算公式所得之乾穀淨重預測值精確。 |
| 英文摘要 | Feed-forward neural network models and linear regression models were developed to estimate the net weight of rice grain after drying and sorting process. A three-layer (one input, one hidden, and one output layers) feed- forward neural network was proposed. There are twoneurons in the input layer for two input data, moisture content and bulk density of rice grain. One neuron is in the hidden layer. And, there is one neuron in the output layer for one output data, net weight of rice grain. The back-propagation learning algorithm was used to minimize the errors. Both linear and quadratic equations had been developed in the linear regression model. According to this study, the average relative errors of estimated data with the neural net model and linear regression models were within2.0%； the neural network model with linear activat function can be considered as a linear regression model. It had been shown the models developed in this study that considered both moisture content and bulk density of rice grain to be more accurate than the method used in the local agricultural organization that only considered moisture content of rice grain. |
本系統中英文摘要資訊取自各篇刊載內容。