how to analysis the heatmap to find the correlation

+1 vote
Sep 28, 2019 in Machine Learning by Vikas
• 130 points
321 views

1 answer to this question.

0 votes

Hi @Vikas, there are 5 simple steps to analyze the heatmap correlation:

1. Import data

data = pd.read_csv('file_clean.csv')

2. Create correlation matrix. .corr() is used to create the correlation matrix. You'll have to make sure that all the elements in the matrix are of numeric type. If they are not of the numeric type you'll have to add or concat them explicitly.

corr = data.corr()

3. Create heatmap in seaborn:

ax = sns.heatmap(
    corr, 
    vmin=-1, vmax=1, center=0,
    cmap=sns.diverging_palette(20, 220, n=200),
    square=True
)
ax.set_xticklabels(
    ax.get_xticklabels(),
    rotation=45,
    horizontalalignment='right'
);

You'll see something like this where the blue indicates positive and red indicates negative. 

Now to start analyzing the heatmap correlation, ask yourself this question:

What's the weakest and strongest correlation pair?

I am assuming its difficult to analyze right? 

Now according to your dataset, you need to create a scatter plot which makes it easier to analyze.

def heatmap(x, y, size):
    fig, ax = plt.subplots()
    
    # Mapping from column names to integer coordinates
    x_labels = [v for v in sorted(x.unique())]
    y_labels = [v for v in sorted(y.unique())]
    x_to_num = {p[1]:p[0] for p in enumerate(x_labels)} 
    y_to_num = {p[1]:p[0] for p in enumerate(y_labels)} 
    
    size_scale = 500
    ax.scatter(
        x=x.map(x_to_num), # Use mapping for x
        y=y.map(y_to_num), # Use mapping for y
        s=size * size_scale, # Vector of square sizes, proportional to size parameter
        marker='s' # Use square as scatterplot marker
    )
    
    # Show column labels on the axes
    ax.set_xticks([x_to_num[v] for v in x_labels])
    ax.set_xticklabels(x_labels, rotation=45, horizontalalignment='right')
    ax.set_yticks([y_to_num[v] for v in y_labels])
    ax.set_yticklabels(y_labels)
    
data = pd.read_csv('https://raw.githubusercontent.com/drazenz/heatmap/master/autos.clean.csv')
columns = ['bore', 'stroke', 'compression-ratio', 'horsepower', 'city-mpg', 'price'] 
corr = data[columns].corr()
corr = pd.melt(corr.reset_index(), id_vars='index') # Unpivot the dataframe, so we can get pair of arrays for x and y
corr.columns = ['x', 'y', 'value']
heatmap(
    x=corr['x'],
    y=corr['y'],
    size=corr['value'].abs()
)

You'll get something like this:

Make a few modifications(get the plots in between the grid)

ax.grid(False, 'major')
ax.grid(True, 'minor')
ax.set_xticks([t + 0.5 for t in ax.get_xticks()], minor=True)
ax.set_yticks([t + 0.5 for t in ax.get_yticks()], minor=True)

And you are good to go!
Have a look at this blog: https://towardsdatascience.com/better-heatmaps-and-correlation-matrix-plots-in-python-41445d0f2bec

answered Sep 30, 2019 by Vishal

Related Questions In Machine Learning

0 votes
1 answer
0 votes
0 answers

How to know if a problem is solvable by machine learning?

I have recently started learning the machine. ...READ MORE

Nov 21, 2019 in Machine Learning by Hannah
• 17,330 points
49 views
0 votes
1 answer

Example to run KNN algorithm using python

Have a look at this one: from sklearn.datasets ...READ MORE

answered May 8, 2019 in Machine Learning by Harvy
212 views
0 votes
1 answer
0 votes
1 answer
0 votes
1 answer

What is correlation and its types?

Correlation is a statistical measure that shows ...READ MORE

answered May 9, 2019 in Machine Learning by Zulaikha
96 views
0 votes
1 answer

How do I create a decision tree?

Let us consider the following example. Suppose a ...READ MORE

answered May 13, 2019 in Machine Learning by Fatima
89 views
+1 vote
1 answer

ValueError: Not enough values to unpack

Make the following changes in your script, ...READ MORE

answered Jun 24, 2019 in Machine Learning by Omkar
• 68,840 points
2,920 views