读取具有多个输入和输出的 CSV 文件

Question

I have the following table and would like to find out the relationship between inputs and outputs in order to make predictions.

Later I want enter input values for heater_power, voltage, heater_efficiency and heater_mass and generate a prediction for the outputs.

In the table you can see that I have 4 input parameters and 3 output parameters.

table

I have created a code. The values for input and output are manually written to an array.

Import

import tensorflow as tf
import numpy as np

Set up training data

inputMatrix = np.array([(100,230,0.95,100),
                        (200,245,0.99,121),
                        ( 40,250,0.91,123)],dtype=float)
outputMatrix = np.array([(120, 5,120),
                         (123,24,100),
                         (154, 3,121)],dtype=float)
for i,c in enumerate(inputMatrix):
print("{}Input Matrix={}Output Matrix".format(c,outputMatrix[i]))

Create the Model

l0 = tf.keras.layers.Dense(units = 4, input_shape = [4])
l1 = tf.keras.layers.Dense(units = 64)
l2 = tf.keras.layers.Dense(units = 128)
l3 = tf.keras.layers.Dense(units = 3)

model = tf.keras.Sequential([l0,l1,l2,l3])

Compile the Model

model.compile(loss='mean_squared_error', optimizer=tf.keras.optimizers.Adam(0.1))

Train the model

history = model.fit(inputMatrix,outputMatrix,epochs=500,verbose=False)
print("Finished training the model!")

Display training statistics

import matplotlib.pyplot as plt
plt.xlabel('Epoch Number')
plt.ylabel('Loss Magnitude')
plt.plot(history.history['loss'])

Use the model to predict values

print(model.predict(np.array([120,260,0.98,110]).reshape(1,4)))

I now want to read the table automatically from a csv file. The data should be separated according to input and output and read in.

How do I do this? Does it make sense to work with arrays here or are there better possibilities?

I have some doubts that my approach is basically correct. My code seems to be so short. Or do I have to choose a different approach for my problem?

Answer 1

You need to load the DataFrame using MultiIndex since you are having both Input, Output, and all the Features in different levels.

Below is the code to do so, I have used column names like A,B,C etc, change it according to your data.

import pandas as pd
import numpy as np 

index = pd.MultiIndex.from_tuples([("Input","B"),("Input","C"),("Input","D"),("Input","E"),("Output","F"),("Output","G"),("Output","H")])
df = pd.read_csv("sample_csv.csv",header=[0,1],index_col=0)
df.columns = index 

df:

df.columns

MultiIndex([( 'Input', 'B'),
            ( 'Input', 'C'),
            ( 'Input', 'D'),
            ( 'Input', 'E'),
            ('Output', 'F'),
            ('Output', 'G'),
            ('Output', 'H')],
           )
input_data = df[[( 'Input', 'B'),
            ( 'Input', 'C'),
            ( 'Input', 'D'),
            ( 'Input', 'E')]].values

input_data = list(map(lambda x:tuple(x),input_data))

#Input_data

[(100.0, 230.0, 0.95, 100.0),
 (200.0, 245.0, 0.99, 121.0),
 (40.0, 250.0, 0.91, 123.0)]



output_data = df[[('Output', 'F'),
            ('Output', 'G'),
            ('Output', 'H')]].values

output_data = list(map(lambda x:tuple(x),output_data))

#Ouput_data

[(120, 5, 120), (123, 24, 100), (154, 3, 121)]