import numpy as np
import matplotlib.pyplot as plt
from sklearn.neural_network import MLPClassifier
from sklearn.datasets import load_iris
from pandas import DataFrame
import pandas as pd
import seaborn as sns
# Input data into dataframe
#
iris = load_iris()
x = iris['data']
#
# make data frame
df = DataFrame(x)
df.columns = ['sepal_length', 'sepal_width', 'petal_length', 'petal_width']
# add column with targets
df['target'] = iris['target']
#
#add species name for plotting
df['species'] = df['target'].map( {0:iris.target_names[0], 1:iris.target_names[1], 2:iris.target_names[2]})
# Create NumPy target array
target = np.array(df.target.values)
# Split data in test and train sets using 70-30 split
#
from sklearn.model_selection import train_test_split
(X_train, X_test, Y_train, Y_test) = train_test_split(x,target,test_size=0.3)
# ANN sensitive to scaling so scale
#
# Import StandardScaler and scale
from sklearn.preprocessing import StandardScaler
scaler = StandardScaler()
scaler.fit(X_train)
X_train = scaler.transform(X_train)
X_test = scaler.transform(X_test)
# Create classifier and train
mlp = MLPClassifier(max_iter=300, solver='adam', random_state=1)
mlp.fit(X_train,Y_train)
#
# Print out precent accuracy of training set
accuracy = round(mlp.score(X_train,Y_train)*100,2)
print (f'% Accuracy = {accuracy}%')
% Accuracy = 98.1%
/shared-libs/python3.7/py/lib/python3.7/site-packages/sklearn/neural_network/_multilayer_perceptron.py:696: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (300) reached and the optimization hasn't converged yet.
ConvergenceWarning,
# Print out percent accuracy of test set
accuracy = round(mlp.score(X_test,Y_test)*100,2)
print (f'Accuracy = {accuracy}%')
Accuracy = 95.56%
# Scale the entire data X
scaler.fit(x)
X = scaler.transform(x)
# Predict the entire dataset and add column to dataframe for Prediction and predicted species
pred = mlp.predict(X)
df['prediction'] = pred
df['pred_species'] = df['prediction'].map( {0:iris.target_names[0], 1:iris.target_names[1], 2:iris.target_names[2]})
# Plot petal length vs sepal length for actual data with hue = species and for prediction; compare
sns.relplot(x='petal_length',y='sepal_length',data=df,hue='species')
sns.relplot(x='petal_length',y='sepal_length',data=df,hue='pred_species')
# Plot petal width vs sepal width for actual data with hue = species and for prediction; compare
sns.relplot(x='petal_width',y='sepal_width',data=df,hue='species')
sns.relplot(x='petal_width',y='sepal_width',data=df,hue='pred_species')
The actual classification has the data much more spread out compared to the predicted.