Seaborn Pairplot with Dataframe vs CSV

Question

I have a dataframe in a Jupyter notebook and do a pairplot on it to get a bunch of plots against each other.

import seaborn as sns
sns.pairplot(df_merge)

Here is the pairplot as a result.

However, it plots the data incorrectly and in a non-aesthetic way. However, when I export this dataframe to a csv and then read it back into the program as a dataframe:

import seaborn as sns
df_merge.to_csv('dataframe.csv')
x = pd.read_csv('dataframe.csv')
sns.pairplot(x)

Sns plots it fine and the correlations between variables can be seen but I have an unnecessary column called Unnamed which I don't need.

Does anyone know what could cause this issue and how I can go about correcting it without needing to export the dataframe as a csv?

Answer 1

When you do:

df_merge.to_csv('dataframe.csv')

you write also the index of df_merge without a name. Then

x = pd.read_csv('dataframe.csv')

reads the index as Unnamed 0 column. To fix this, either save the data frame without index:

df_merge.to_csv('dataframe.csv', index=False)
x = pd.read_csv('dataframe.csv')

or read the csv with index:

df_merge.to_csv('dataframe.csv')
x = pd.read_csv('dataframe.csv', index_col=[0])

Answer 2

Figured out that the issue I was having was when I was changing the dataframe to a CSV and then changing it back to a dataframe, the values in the dataframe had a float64 type where as in my dataframe before they were all objects. Converting all the numerical columns to float before plotting the graph solved my issue.