eponkratova

+--------+-------+-------+-------+-------+----+-------+-------+-------+--+
| Branch |  x1   |  x2   |  x3   |  x4   | x5 |  y1   |  y2   |  y3   |  |
+--------+-------+-------+-------+-------+----+-------+-------+-------+--+
| A      | 86.13 | 16.24 | 48.21 | 49.69 |  9 | 54.53 | 58.98 | 38.16 |  |
| B      | 29.26 | 10.24 | 41.96 | 40.65 |  5 | 24.69 | 33.89 | 26.02 |  |
| C      | 43.12 | 11.31 | 38.19 | 35.03 |  9 | 36.41 | 40.62 | 28.51 |  |
+--------+-------+-------+-------+-------+----+-------+-------+-------+--+

eponkratova

class DEA(object):
    random.seed(5)
    def __init__(self, inputs, outputs):
        """
        Initialize the DEA object with input data
        n = number of entities (observations)
        m = number of inputs (variables, features)
        r = number of outputs
        :param inputs: inputs, n x m numpy array
        :param outputs: outputs, n x r numpy array

eponkratova

    def __efficiency(self, unit):
        """
        Efficiency function with already computed weights
        :param unit: which unit to compute for
        :return: efficiency
        """
        # compute efficiency
        denominator = np.dot(self.inputs, self.input_w)
        numerator = np.dot(self.outputs, self.output_w)
        return (numerator/denominator)[unit]

eponkratova

    def __target(self, x, unit):
        """
        Theta target function for one unit
        :param x: combined weights
        :param unit: which production unit to compute
        :return: theta
        """
        in_w, out_w, lambdas = x[:self.m], x[self.m:(self.m+self.r)], x[(self.m+self.r):]  # unroll the weights
        denominator = np.dot(self.inputs[unit], in_w)
        numerator = np.dot(self.outputs[unit], out_w)

eponkratova

    def __constraints(self, x, unit):
        """
        Constraints for optimization for one unit
        :param x: combined weights
        :param unit: which production unit to compute
        :return: array of constraints
        """
        in_w, out_w, lambdas = x[:self.m], x[self.m:(self.m+self.r)], x[(self.m+self.r):]  # unroll the weights
        constr = []  # init the constraint array
        # for each input, lambdas with inputs

eponkratova

    def __optimize(self):
        """
        Optimization of the DEA model
        Use: http://docs.scipy.org/doc/scipy-0.17.0/reference/generated/scipy.optimize.linprog.html
        A = coefficients in the constraints
        b = rhs of constraints
        c = coefficients of the target function
        :return:
        """
        d0 = self.m + self.r + self.n

eponkratova

    def name_units(self, names):
        """
        Provide names for units for presentation purposes
        :param names: a list of names, equal in length to the number of units
        :return: nothing
        """
        assert(self.n == len(names))
        self.names = names

    def fit(self):

eponkratova

def save_results(dataframe):  
    df_results = pd.DataFrame([]) 
    df_results = df_results.append(not_efficient, ignore_index = True).T
    df_results = df_results.reset_index()
    df_results.columns = ['dmu', 'efficiency']    
    dataframe = dataframe.merge(df_results)
    return dataframe

eponkratova

df_ineff = save_results(df)
df_complete = save_results_complete(df)
df_complete.head()

#Estimating the improvement options - input oriented
heading = list(df_ineff.iloc[:, 1:6])
inter = []
for c in df_ineff.iloc[:, 1:6].columns:
    inter.append(df_ineff[c].multiply((1- df_ineff.iloc[:,-1])))
    df_improvement = round(pd.concat(inter, axis = 1),1)

eponkratova

Getting Started With Superset: Airbnb’s data exploration platform

These instructions are for Amazon Linux Version 2

Update Python and PIP versions on EC2 (Amazon AMI)

sudo yum update -y
sudo yum install python3 -y