Weka 神经网络分析
常用的神经网络就是向前反馈的 BP(Back Propagation) 网络,也叫多层前馈网络,而 BP 在 weka 中就是由 MultilayerPerceptron 算法实现的。
(1) 在 Weka 主界面选择 Explorer>Open file 选择数据文件(将xls转换成csv格式)
(2) 选择 Classify 选项卡,选择 Choose 按钮,然后再选择 functions>MultiLayerPreceptron
(3) 在 Test Option 中选择 Use Training set ,然后单击 Choose 右侧文本框
GUI 下拉框,选择 True ,其他默认,在 Start 按钮上方选择输出层,即要预测的值,然后点击点击 OK 按钮,然后单击 Start 按钮,设置参数,
点击 Start ,运行后选择 Accept 。( momentum 带动量)
得到运行结果,在 Exploerer>Classifier output 中可以看到结果:
=== Run information ===
Scheme: weka.classifiers.functions.MultilayerPerceptron -L 0.3 -M 0.2 -N 500 -V 0 -S 0 -E 20 -H a -G -R
Relation: shuizhi2003-weka.filters.unsupervised.attribute.Remove-R2-7
Instances: 23
Attributes: 7
溶解氧
B1
B2
B3
B4
B5
B6
Test mode: evaluate on training data
=== Classifier model (full training set) ===
Linear Node 0
Inputs Weights
Threshold -0.7803035779165746
Node 1 -3.5991023803526447
Node 2 2.1968777991012516
Node 3 3.146316946815286
Sigmoid Node 1
Inputs Weights
Threshold -2.1644039121288583
Attrib B1 3.5316131171975034
Attrib B2 -3.2085377318533648
Attrib B3 -3.040826618503377
Attrib B4 2.5982786224902887
Attrib B5 -0.11957884133740122
Attrib B6 1.844990948707308
Sigmoid Node 2
Inputs Weights
Threshold -1.0555748116838761
Attrib B1 -0.3887415141267751
Attrib B2 1.7835144295128773
Attrib B3 -1.404906253700419
Attrib B4 -2.458278666190213
Attrib B5 4.483492439027104
Attrib B6 -4.021881778908347
Sigmoid Node 3
Inputs Weights
Threshold -0.4134171735223275
Attrib B1 0.12309277936284704
Attrib B2 1.278817931602191
Attrib B3 0.5473417036767106
Attrib B4 0.18852178888400642
Attrib B5 3.3505018112173235
Attrib B6 2.1259296458835206
Class
Input
Node 0
Time taken to build model: 319.32 seconds
=== Evaluation on training set ===
=== Summary ===
Correlation coefficient 0.9261
Mean absolute error 0.9066
Root mean squared error 1.1904
Relative absolute error 34.8133 %
Root relative squared error 40.4467 %
Total Number of Instances 23
0
顶