[FIXED] Dataset indices for predicted values is not matching with those for actual values

Issue

I am a python novice who is trying to solve a regression problem with neural networks. I am at the stage where I want to plot the predicted vs actual followed by determining the regression coefficient.

Model training

#import statements 

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from sklearn.neural_network import MLPRegressor
from sklearn.model_selection import train_test_split
%matplotlib inline

#importing the dataset
data = pd.read_csv("PPV_dataset.csv")
X = np.array(data.drop(["PPV"],1))
y = np.array(data["PPV"])

#model training & prediction
nn = MLPRegressor(hidden_layer_sizes=(100,), activation = 'logistic', solver = 'sgd')
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.25)
nn.fit(X_train, y_train)
pred = nn.predict(X_test)

#indices of test set
a = X_test
indices = []
for row in range(len(X)):
    for i in range(len(a)):
        if np.all(a[i]==X[row]):
            indices.append(row)

#listing actual values in an array
actual_values = []
for i in range(len(indices)):
    actual_values.append(y[indices[i]])

Comparing actual to predicted values

len(actual_values)
13

len(pred)
12

Image of dataset

Solution

You should use the matplotlib and the seaborn libraries for plotting you graph, and for coeficient r_sq = nn.score(actual_values, pred)

I recommend using seaborn.lmplot() in your case

for roberts particular case I suggest:

from sklearn.metrics import r2_score
r2_score(y_true, y_pred)

Answered By - Divyessh