numpy数组转换错误

Question

I have a dataset with string and float data. numPy tries to convert everything to a float, giving the error "cannot convert string to float"

import numpy as np
import scipy
import matplotlib.pyplot as plt
import pandas as pd
from sklearn import datasets
from sklearn.model_selection import train_test_split
from sklearn.neighbors import KNeighborsClassifier

pd.set_option('display.height', 750)
pd.set_option('display.width', 750)

colnames = ['AGE', 'WORKCLASS', 'FNLWGT','EDU','EDU-NUM','MARITAL- 
STATUS','JOB','RELATIONSHIP','RACE', 'SEX', 'CAPITAL-GAIN', 'CAPITAL- 
LOSS','HOURS-PER-WEEK', 'NATIVE-COUNTRY', 'INCOME']
url = 'https://archive.ics.uci.edu/ml/machine-learning- 
databases/adult/adult.data'
adults = pd.read_csv(url, names=colnames, header=None)

adults['CAPITAL-GAINS'] = (adults['CAPITAL-GAIN'] - adults['CAPITAL-LOSS'])

adults = adults.drop(['RELATIONSHIP', 'FNLWGT', 'EDU-NUM', 'MARITAL-STATUS', 
'CAPITAL-GAIN', 'CAPITAL-LOSS'], axis=1)
#rearrange the columns to make it easier to set X
adults = adults[['AGE', 'WORKCLASS','EDU','JOB','RACE', 'SEX','HOURS-PER- 
WEEK', 'NATIVE-COUNTRY', 'CAPITAL-GAINS', 'INCOME']]
adults.replace({'?': 0}, inplace=True)
#assign the X and y arrays using numpy
X = np.array(adults.ix[:,0:9])
y = np.array(adults['INCOME'])

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.25)
knn = KNeighborsClassifier()
knn.fit(X_train ,y_train)
pred = knn.predict(X_test)
print (accuracy_score(y_test, pred))

traceback:

Traceback (most recent call last):
  File "C:/Users/nolan/OneDrive/Desktop/digits.py", line 37, in <module>
    knn.fit(X_train ,y_train)
  File "C:\Program Files\Python\lib\site-packages\sklearn\neighbors\base.py", line 765, in fit
    X, y = check_X_y(X, y, "csr", multi_output=True)
  File "C:\Program Files\Python\lib\site-packages\sklearn\utils\validation.py", line 573, in check_X_y
    ensure_min_features, warn_on_dtype, estimator)
  File "C:\Program Files\Python\lib\site-packages\sklearn\utils\validation.py", line 433, in check_array
    array = np.array(array, dtype=dtype, order=order, copy=copy)
ValueError: could not convert string to float: ' Peru'

all the data looks like this:

39, State-gov, 77516, Bachelors, 13, Never-married, Adm-clerical, Not-in-family, White, Male, 2174, 0

is there a way to set numPy to hold this data with the conversion error?

Answer 1

There is not any numpy conversion error here; the issue is simply than the k-nn algorithm cannot handle categorical features. It is true that this is not explicitly mentioned in the scikit-learn documentation , but it follows directly if you have even a rough idea of what the algorithm does, which is computing distances between the data points, so that it can subsequently find the k nearest ones, hence the name. And since there is not any (simple & general) way to compute distances between categorical features, the algorithm is simply not applicable in such cases.

See also this answer at Data Science Stack Exchange.

Answer 2

you should change the classifier, if possible. SVM and neural networks support this type of data, but KNN not suport this.