MAPE calculation in python

Question

I want to calculate Mean Absolute percentage error (MAPE) of predicted and true values. I found a solution from here, but this gives error and shows invalid syntax in the line mask = a <> 0

    def mape_vectorized_v2(a, b): 
    mask = a <> 0
    return (np.fabs(a - b)/a)[mask].mean() 

   def mape_vectorized_v2(a, b): 
       File "<ipython-input-5-afa5c1162e83>", line 1
         def mape_vectorized_v2(a, b):
                                       ^
     SyntaxError: unexpected EOF while parsing

I am using spyder3. My predicted value is a type np.array and true value is dataframe

type(predicted)
Out[7]: numpy.ndarray
type(y_test)
Out[8]: pandas.core.frame.DataFrame

How do i clear this error and proceed with MAPE Calculation ?

Edit :

predicted.head()
Out[22]: 
   Total_kWh
0   7.163627
1   6.584960
2   6.638057
3   7.785487
4   6.994427

y_test.head()
Out[23]: 
     Total_kWh
79         7.2
148        6.7
143        6.7
189        7.2
17         6.4

np.abs(y_test[['Total_kWh']] - predicted[['Total_kWh']]).head()
Out[24]: 
   Total_kWh
0        NaN
1        NaN
2        NaN
3        NaN
4   0.094427

Answer 1

In python for compare by not equal need !=, not <>.

So need:

def mape_vectorized_v2(a, b): 
    mask = a != 0
    return (np.fabs(a - b)/a)[mask].mean()

Another solution from stats.stackexchange:

def mean_absolute_percentage_error(y_true, y_pred): 
    y_true, y_pred = np.array(y_true), np.array(y_pred)
    return np.mean(np.abs((y_true - y_pred) / y_true)) * 100

Answer 2

The new version of scikit-learn (v0.24) has a function that will calculate MAPE. sklearn.metrics.mean_absolute_percentage_error

All what you need is two array-like variables: y_true storing the actual/real values, and y_pred storing the predicted values.

You can refer to the official documentation here.

Answer 3

Both solutions are not working with zero values. This is working form me:

def percentage_error(actual, predicted):
    res = np.empty(actual.shape)
    for j in range(actual.shape[0]):
        if actual[j] != 0:
            res[j] = (actual[j] - predicted[j]) / actual[j]
        else:
            res[j] = predicted[j] / np.mean(actual)
    return res

def mean_absolute_percentage_error(y_true, y_pred): 
    return np.mean(np.abs(percentage_error(np.asarray(y_true), np.asarray(y_pred)))) * 100

I hope it helps.