Source code for mlreflect.training.prediction

from typing import Union, List

import numpy as np
import tensorflow.keras as keras
from numpy import ndarray
from pandas import DataFrame
from tensorflow.keras import Model

from ..utils.label_helpers import convert_to_dataframe
from ..utils.performance_tools import timer


class Prediction:
    def __init__(self, model_path: str, label_names: List[str]):
        self.model_path = model_path
        self.model = self._load_model_from_file(model_path)

        self.label_names = label_names

    @timer
    def predict_labels(self, test_input: ndarray):
        try:
            test_input = np.asarray(test_input)
        except TypeError:
            raise TypeError('test_input must be castable to ndarray')

        test_input = np.atleast_2d(test_input)

        predicted_labels = self.model.predict(test_input)
        predicted_labels = convert_to_dataframe(predicted_labels, self.label_names)

        return predicted_labels

    def mean_absolute_percentage_error(self, predicted_labels: Union[DataFrame, ndarray],
                                       test_labels: Union[DataFrame, ndarray]):

        test_labels = convert_to_dataframe(test_labels, self.label_names)
        predicted_labels = convert_to_dataframe(predicted_labels, self.label_names)

        absolute_percentage_error = abs(test_labels.reset_index() - predicted_labels.reset_index()) / abs(
            test_labels.reset_index())
        del absolute_percentage_error['index']
        mean_absolute_percentage_error = absolute_percentage_error.mean()

        return mean_absolute_percentage_error

    def mean_absolute_error(self, predicted_labels: Union[DataFrame, ndarray], test_labels: Union[DataFrame, ndarray]):

        test_labels = convert_to_dataframe(test_labels, self.label_names)
        predicted_labels = convert_to_dataframe(predicted_labels, self.label_names)

        absolute_error = abs(test_labels.reset_index() - predicted_labels.reset_index())
        del absolute_error['index']
        mean_absolute_error = absolute_error.mean()

        return mean_absolute_error

    @staticmethod
    def _load_model_from_file(model_path: str) -> Model:
        return keras.models.load_model(model_path)

    @staticmethod
    def _wrap_ndarray_in_list(test_input: Union[ndarray, List[ndarray]]):
        if type(test_input) is ndarray:
            return [test_input]
        elif type(test_input) is list:
            return test_input
        else:
            raise TypeError('test_input must be ndarray or list of ndarrays.')