0
$\begingroup$

I am working with typing data with timing features(unit: ms) and some of the features are based on the keyboard keyCodes(positive integers, range:[8, 222]). Currently, I use StandardScaler() by scikit-learn to scale all the features, so that my learning models do not overweigh the keyCode based features. I would like dicretize the keyCode based features and run the StandardScaler() for the timing features only. How can I go about this?

$\endgroup$
0
$\begingroup$

Not sure about your question, but maybe something like the following could help:

  1. Create dummy variables for your KeyCodes variable
  2. Normalize only those variables using the StandardScaler.
import pandas as pd
import random
from sklearn.preprocessing import StandardScaler

# Let's assume the following dataframe
data = {
    'KeyCodes': [random.randrange(2, 223, 1) for _ in range(10000)],
    'age': [random.randrange(1, 100) for _ in range(10000)],
    'id': [i for i in range(10000)]
}

# First you need to create the dummy variables based in KeyCodes
df = pd.DataFrame.from_dict(data)
df.head()
dummies = pd.get_dummies(df['KeyCodes']).rename(columns=lambda x: f'KeyCode_{x}')
df = pd.concat([df, dummies], axis=1)
df.drop(['KeyCodes'], inplace=True, axis=1)

# Then you can apply the normalization to the subset of features you wish to normalize
normalized_df = df.copy()
col_names = [f'KeyCode_{i}' for i in range(8, 223)]
features_normalized = df[col_names]
scaler = StandardScaler().fit(features_normalized.values)
features_normalized = scaler.transform(features_normalized.values)
normalized_df[col_names] = features_normalized

# Explore the output
normalized_df['age'][:10]  # you can see it was not normalized
set(normalized_df['KeyCodes_8'])  # normalized version of the feature
set(df['KeyCodes_8'])  # not normalized version of the feature
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.