ValueError: could not convert string to float: med

Question

I am writing a very simple script. All I have to do is read data using panda and then train a decision tree on data. Data that I am using is:

https://archive.ics.uci.edu/ml/machine-learning-databases/car/car.data

And following is my script

import numpy as np
from sklearn.cross_validation import train_test_split
from sklearn.tree import DecisionTreeClassifier
from sklearn.metrics import accuracy_score
from sklearn import tree
from sklearn import preprocessing
import pandas as pd
balance_data=pd.read_csv("https://archive.ics.uci.edu/ml/machine-learning-databases/car/car.data",
                           sep= ',', header= None)
#print "Dataset:: "

#df1.head()

X = balance_data.values[:, 0:5]
Y = balance_data.values[:,6]
X_train, X_test, y_train, y_test = train_test_split( X, Y, test_size = 0.2, random_state = 100)
clf_gini = DecisionTreeClassifier(criterion = "gini", random_state = 100,
                               max_depth=3, min_samples_leaf=5)

clf_gini.fit(X_train, y_train)

From the error I am guessing that it couldn't convert "med" attribute value to float. And by looking at the data my random guess is that low has a space before it and med doesn't. That is why it is getting confused. But I am not sure of it. Please tell what could be wrong with it. PS: error is occurring at the last line and here is the traceback

ValueError                                Traceback (most recent call last)
<ipython-input-26-b495e5f26174> in <module>()
     18                                max_depth=3, min_samples_leaf=5)
     19 X_train[X_train != '']
---> 20 clf_gini.fit(X_train, y_train)

/home/fatima/anaconda2/lib/python2.7/site-packages/sklearn/tree/tree.pyc in fit(self, X, y, sample_weight, check_input, X_idx_sorted)
    788             sample_weight=sample_weight,
    789             check_input=check_input,
--> 790             X_idx_sorted=X_idx_sorted)
    791         return self
    792 

/home/fatima/anaconda2/lib/python2.7/site-packages/sklearn/tree/tree.pyc in fit(self, X, y, sample_weight, check_input, X_idx_sorted)
    114         random_state = check_random_state(self.random_state)
    115         if check_input:
--> 116             X = check_array(X, dtype=DTYPE, accept_sparse="csc")
    117             y = check_array(y, ensure_2d=False, dtype=None)
    118             if issparse(X):

/home/fatima/anaconda2/lib/python2.7/site-packages/sklearn/utils/validation.pyc in check_array(array, accept_sparse, dtype, order, copy, force_all_finite, ensure_2d, allow_nd, ensure_min_samples, ensure_min_features, warn_on_dtype, estimator)
    400                                       force_all_finite)
    401     else:
--> 402         array = np.array(array, dtype=dtype, order=order, copy=copy)
    403 
    404         if ensure_2d:

ValueError: could not convert string to float: med

Answer 1

The dataset looks like this:

       0      1  2  3      4     5      6
0  vhigh  vhigh  2  2  small   low  unacc
1  vhigh  vhigh  2  2  small   med  unacc
2  vhigh  vhigh  2  2  small  high  unacc
3  vhigh  vhigh  2  2    med   low  unacc
4  vhigh  vhigh  2  2    med   med  unacc

Where the data types (dtypes) are all objects. However, machine learning algorithms can only learn from numbers (int, float, doubles .. ) thus, you need to encode your data before you use it for training.

There are several ways to encode your data, one way is to use label encoding, to do that, add the following lines to your code just after loading the dataset:

le = preprocessing.LabelEncoder()
balance_data = balance_data.apply(le.fit_transform)

Now the data in balance_data looks like this:

   0  1  2  3  4  5  6
0  3  3  0  0  2  1  2
1  3  3  0  0  2  2  2
2  3  3  0  0  2  0  2
3  3  3  0  0  1  1  2
4  3  3  0  0  1  2  2

where all data types are int.

In general, you need to perform some data preprocessing before training/fitting your model. For that, I recommend that you go through some tutorial to understand the process. For instance, check this:

• practical-guide-data-preprocessing-python-scikit-learn

---

Here's the overall code with the fix:

import numpy as np
from sklearn.cross_validation import train_test_split
from sklearn.tree import DecisionTreeClassifier
from sklearn.metrics import accuracy_score
from sklearn import tree
from sklearn import preprocessing
import pandas as pd
balance_data=pd.read_csv("https://archive.ics.uci.edu/ml/machine-learning-databases/car/car.data",
                           sep= ',', header= None)
#print "Dataset:: "

#df1.head()

le = preprocessing.LabelEncoder()
balance_data = balance_data.apply(le.fit_transform)

X = balance_data.values[:, 0:5]
Y = balance_data.values[:,6]
X_train, X_test, y_train, y_test = train_test_split( X, Y, test_size = 0.2, random_state = 100)
clf_gini = DecisionTreeClassifier(criterion = "gini", random_state = 100,
                               max_depth=3, min_samples_leaf=5)

clf_gini.fit(X_train, y_train)

Answer 2

I've checked the file you are trying to process, and I found this is the data:

vhigh,vhigh,2,2,small,med,unacc
vhigh,vhigh,2,2,small,high,unacc
vhigh,vhigh,2,2,med,low,unacc
vhigh,vhigh,2,2,med,med,unacc
vhigh,vhigh,2,2,med,high,unacc

So when you ask for train the model, internally is trying to convert your vector into numbers, but founds strings values (as "small", "med", "high", etc) wich are not parseables to a number.

A good start could be normalize your categorical values into onehot encoding. Check it here:

http://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.OneHotEncoder.html