整体处理的想法是
首先加载训练数据，训练数据后进行预测，最后输出预测值。
主() )
importnumpyaspyimportdataloadasdataloadasdataloadimportmodeltrainasmodeltrainimportdatapredictionasdatapredictiondefmain (3360数据_ Tata (#训练数据print(data_train ) modeltrain.train_model ) data _ pre=data load .训练加载数据加载需要预测的数据的data _ real=data load.load _ data _ real ) #要加载的数据的真值print(3) ) ) ) 65 pre _ result
importnumpyasnpimportpandasaspddata=PD.read _ CSV ('./2015 ) z ).CSV ' (读取CSV格式文件中的所有数据的data1=NP.mat ) (data ) )数据矩阵化print(data1 ) def load_data_train ) 320 培训数据print(data_train ) return data _ traindefload _ data _ pre ) 3360 data _ pre=data _ train 列是输入数据data_mean=data_pre.mean ) #求平均值data_std=data_pre.std的data_std #标准化return data _ predefload _ DDD
importnumpyasnpfromkeras.modelsimportsequentialfromkeras.layers.coreimportdense， activationdeftrain _ model (data _ train ) : model file='./model weight ' #训练模型中的权重y _ mean _ STD='./data_train=NP.matrix (data _ train ).astype('float64 ' ) data_mean=NP.mean ) data _ train， axis=0)。axis=0) #每一列的标准偏差# data _ train=(data _ train-data _ mean (/data _ STD print )1) x _ train=datd print -1) )输入除最后一列之外的所有数据x y_train=data_train[:data_train.shape[1] - 1] #输出所有数据的最后一列yprint(x_train 模型训练模型=sequential ) )作为model.add的input_dim=x_train.shape[1]，kernel_initializer='uniform ' ) input_dim=x_train.shape[1] ) (model.com pile (loss=' mean _ squared _ eror '，optimizer='adam ' ) ) MP batch _ size=x _ train.shape [0] (model.save _ weights (model file ) #保存模型权重y_mean=data_mean ) : data data _ train.shape [1]-1 ] print ('训练完成' )将标准差中的参数写入文件f=open(y_mean_STD， ' w ' ) mean_STD=str(y_mean.astype(str ) ) str (y _ STD.as type (str ) ) mean _ STD=mean _ STD.rep Pepe
importnumpyasnpfromkeras.modelsimportsequentialfromkeras.layers.coreimportdense， activationdefload _ y (3360 f=oppy ' r ') y _ mean=f.read (y _ mean=y _ mean.split ) ' ) f.close ) returny， input_dim=data_pre.shape[1]， kernel_initializer='uniform ' ) (print(data_pre.shape[1] ) model.add ) activation(relu ' ) model.add model.com pile (loss=' mean _ squared _ error '， optimizer='adam ' ) model.load _ weights ('./model weight ' ) mean_std=load_y ) pre _ result=model.pre esess