Scaling data and KNN Regression

Machine Learning for Finance in Python

Nathan George

Data Science Professor

feature importances

Feature selection: remove weekdays

print(feature_names)

['10d_close_pct',
 '14-day SMA',
 '14-day RSI',
 '200-day SMA',
 '200-day RSI',
 'Adj_Volume_1d_change',
 'Adj_Volume_1d_change_SMA',
 'weekday_1',
 'weekday_2',
 'weekday_3',
 'weekday_4']

print(feature_names[:-4])

['10d_close_pct',
 '14-day SMA',
 '14-day RSI',
 '200-day SMA',
 '200-day RSI',
 'Adj_Volume_1d_change',
 'Adj_Volume_1d_change_SMA']

Remove weekdays

train_features = train_features.iloc[:, :-4]
test_features = test_features.iloc[:, :-4]

2D feature plot

knn prediction unknown

knn prediction with 2 nearest points

minowski distance

large and small feature

Scaling options

Scaling options:

min-max
standardization
median-MAD
map to arbitrary function (e.g. sigmoid, tanh)

2D feature plots before and after scaling

sklearn's scale

from sklearn.preprocessing import scale

sc = scale()
scaled_train_features = sc.fit_transform(train_features)
scaled_test_features = sc.transform(test_features)

before and after standardization

Making subplots

# create figure and list containing axes
f, ax = plt.subplots(nrows=2, ncols=1)

# plot histograms of before and after scaling
train_features.iloc[:, 2].hist(ax=ax[0])
ax[1].hist(scaled_train_features[:, 2])
plt.show()

Scale data and use KNN!

Machine Learning for Finance in Python