__init__.py

import pandas as pd
import matplotlib.pyplot as plt
import numpy as np
from scipy.stats import spearmanr as sr
from scipy.cluster import hierarchy as hc
from typing import List, Any, Union, Tuple, Optional, Dict
import random, math
# TODO: Remove Dependencies, starting with Sklearn
from sklearn.metrics import roc_curve, \
        precision_recall_curve, roc_auc_score, \
        confusion_matrix
import itertools
import logging

# TODO: Make categorical_cols optional argument (None) to 
# avoid ambiguity when there are no categorical cols


def normalize_numeric(
        df, 
        numerical_cols: List[str] = []):
    """
    Normalizes numeric columns by substracting the mean and dividing
        by standard deviation. If the parameter numerical_cols is not
        provided, it will take all the columns of dtype np.number.

    :Example:

    norm_df = xai.normalize_numeric(
                df,
                normalize_numeric=["age", "other_numeric_attribute"])

    :param df: Pandas Dataframe containing data (inputs and target)
    :type df: pd.DataFrame
    :param numerical_cols: List of strings containing numercial cols
    :type categorical_cols: str
    :returns: Dataframe with normalized numerical values.
    :rtype: pandas.DataFrame

    """
    tmp_df = df.copy()

    if not len(numerical_cols):
        numerical_cols = df.select_dtypes(include=[np.number]).columns

    for k in numerical_cols:
        tmp_df[k] = tmp_df[k].astype(np.float32)
        tmp_df[k] -= tmp_df[k].mean()
        tmp_df[k] /= tmp_df[k].std()

    return tmp_df

def convert_categories(
        df,
        categorical_cols: List[str] = []):
    """
    Converts columns to numeric categories. If the categorical_cols
        parameter is passed as a list then those columns are converted.
        Otherwise, all np.object columns are converted.

    :Example:

    import xai
    cat_df = xai.convert_categories(df)

    :param df: Pandas Dataframe containing data (inputs and target)
    :type df: pandas.DataFrame
    :param categorical_cols: List of strings containing categorical cols
    :type categorical_cols: str
    :returns: Dataframe with categorical numerical values.
    :rtype: pandas.DataFrame

    """
    tmp_df = df.copy()

    if not len(categorical_cols):
        categorical_cols = df.select_dtypes(include=[np.object, np.bool]).columns

    tmp_df[categorical_cols] = tmp_df[categorical_cols].astype('category')
    tmp_df[categorical_cols] = tmp_df[categorical_cols].apply(lambda x: x.cat.codes)
    tmp_df[categorical_cols] = tmp_df[categorical_cols].astype('int8')

    return tmp_df

def group_by_columns(
        df: pd.DataFrame,
        columns: List[str], 
        bins: int = 6,
        categorical_cols: List[str] = []):
    """
    Groups dataframe by the categories (or bucketized values) for all columns provided. 
        If categorical it uses categories,
        if numeric, it uses bins. If more than one column is provided, the columns
        provided are, for example, age and binary_target_label, then the result 
        would be a pandas DataFrame that is grouped by age groups for each of the
        positive and negative/positive labels.

    :Example:

    columns=["loan", "gender"]
    df_groups = xai.group_by_columns(
        df, 
        columns=columns,
        bins=10,
        categorical_cols=["gender", "loan"])

    for group, df_group in df_groups:
        print(group) 
        print(grouped_df.head())

    :param df: Pandas Dataframe containing data (inputs and target)
    :type df: pandas.DataFrame
    :param bins: [Default: 6] Number of bins to be used for numerical cols
    :type bins: int
    :param categorical_cols: [Default: []] Columns within dataframe that are
        categorical. Columns that are not np.objects or np.bool and 
        are not part explicitly
        provided here will be treated as numeric, and bins will be used.
    :type categorical_cols: List[str]
    :returns: Dataframe with categorical numerical values.
    :rtype: pandas.core.groupby.groupby.DataFrameGroupBy

    """

    if not len(categorical_cols):
        categorical_cols = _infer_categorical(df)
    
    group_list = []
    for c in columns:
        col = df[c]
        if c in categorical_cols or not bins:
            grp = c
        else:
            col_min = col.min()
            col_max = col.max()
            # TODO: Use the original bins for display purposes as they may come normalised
            col_bins = pd.cut(col, list(np.linspace(col_min, col_max, bins)))
            grp = col_bins

        group_list.append(grp)

    grouped = df.groupby(group_list)
    return grouped 


def imbalance_plot( 
        df: pd.DataFrame, 
        *cross_cols: str, 
        categorical_cols: List[str] = [],
        bins: int = 6, 
        threshold: float = 0.5):
    """
    Shows the number of examples provided for each of the values across the
        product tuples in the columns provided. If you would like to do processing
        with the sub-groups created by this class please see the 
        group_by_columns function.

    :Example:

    import xai
    class_counts = xai.imbalance_plot(
        df, 
        "gender", "loan",
        bins=10,
        threshold=0.8)

    :param df: Pandas Dataframe containing data (inputs and target)
    :type df: pandas.DataFrame
    :param *cross_cols: One or more positional arguments (passed as *args) that 
        are used to split the data into the cross product of their values 
    :type cross_cross: List[str]
    :param categorical_cols: [Default: []] Columns within dataframe that are
        categorical. Columns that are not np.objects and are not part explicitly
        provided here will be treated as numeric, and bins will be used.
    :type categorical_cols: List[str]
    :param bins: [Default: 6] Number of bins to be used for numerical cols
    :type bins: int
    :param threshold: [Default: 0.5] Threshold to display in the chart.
    :type threshold: float
    :returns: Null
    :rtype: None

    """

    if not cross_cols:
        raise TypeError("imbalance_plot requires at least 1 string column name")

    grouped = group_by_columns(
            df,
            list(cross_cols), 
            bins=bins,
            categorical_cols=categorical_cols)

    grouped_col = grouped[cross_cols[0]]
    count_grp = grouped_col.count()
    count_max = count_grp.values.max()
    ratios = round(count_grp/count_max,4)
    # TODO: Make threshold a minimum number of examples per class
    imbalances = ratios < threshold
    
    cm = plt.cm.get_cmap('RdYlBu_r')
    colors = [cm(1-r/threshold/2) if t else cm(0) \
               for r,t in zip(ratios, imbalances)]
    ax = count_grp.plot.bar(color=colors)
    lp = plt.axhline(threshold*count_max, color='r')
    lp.set_label(f"Threshold: {threshold*count_max:.2f} ({threshold*100:.2f}%)")
    plt.legend()
    plt.show()


def balance(
        df: pd.DataFrame,
        *cross_cols: str, 
        upsample: float = 0.5,
        downsample: int = 1,
        bins: int = 6,
        categorical_cols: List[str] = [],
        plot: bool = True):
    """
    Balances a dataframe based on the columns and cross columns provided.
        The results can be upsampled or downsampled. By default, there is no
        downsample, and the upsample is towards a minimum of 50% of the 
        frequency of the highest class.

    :Example:

    cat_df = xai.balance(
        df, 
        "gender", "loan",
        upsample=0.8,
        downsample=0.8)

    :param df: Pandas Dataframe containing data (inputs and target )
    :type df: pandas.DataFrame
    :param *cross_cols: One or more positional arguments (passed as *args) that 
    are used to split the data into the cross product of their values 
    :type cross_cols: List[str]
    :param upsample: [Default: 0.5] Target upsample for columns lower 
        than percentage.
    :type upsample: float
    :param downsample: [Default: 1] Target downsample for columns higher 
        than percentage.
    :type downsample: float
    :param bins: [Default: 6] Number of bins to be used for numerical cols
    :type bins: int
    :param categorical_cols: [Default: []] Columns within dataframe that are
        categorical. Columns that are not np.objects and are not part explicitly
        provided here will be treated as numeric, and bins will be used.
    :type categorical_cols: List[str]
    :param threshold: [Default: 0.5] Threshold to display in the chart.
    :type threshold: float
    :returns: Dataframe with categorical numerical values.
    :rtype: pandas.DataFrame

    """

    if not len(categorical_cols):
        categorical_cols = df.select_dtypes(include=[np.object, np.bool]).columns

    grouped = group_by_columns(
                df,
                list(cross_cols), 
                bins=bins,
                categorical_cols=categorical_cols)

    count_grp = grouped.count()
    count_max = count_grp.values.max()
    count_upsample = int(upsample*count_max)
    count_downsample = int(downsample*count_max)

    def norm(x):
        if x.shape[0] < count_upsample:
            return x.sample(count_upsample, replace=True)
        elif x.shape[0] > count_downsample:
            return x.sample(count_downsample)
        else:
            return x

    tmp_df = grouped.apply(norm) \
                .reset_index(drop=True)

    if plot:
        imbalance_plot(
            tmp_df,
            *cross_cols,
            bins=bins,
            categorical_cols=categorical_cols)

    return tmp_df
    
def _plot_correlation_dendogram(
        corr: pd.DataFrame, 
        cols: List[str],
        plt_kwargs={}):
    """
    Plot dendogram of a correlation matrix, using the columns provided. 
    This consists of a chart that that shows hierarchically the variables
    that are most correlated by the connecting trees. The closer to the right
    that the connection is, the more correlated the features are.
    If you would like to visualise this as a tree, please 
    see the function _plot_correlation_dendogram.

    :Example:

    columns_to_include=["age", "loan", "gender"]
    xai._plot_correlation_dendogram(df, cols=columns_to_include)

    :returns: Null
    :rtype: None

    """

    corr = np.round(corr, 4)
    corr_condensed = hc.distance.squareform(1-corr)
    z = hc.linkage(corr_condensed, method="average")
    fig = plt.figure(**plt_kwargs)
    dendrogram = hc.dendrogram(
        z, labels=cols, orientation="left", leaf_font_size=16)
    plt.show()

def _plot_correlation_matrix(
        corr, 
        cols: List[str], 
        plt_kwargs={}):
    """
    Plot a matrix of the correlation matrix, using the columns provided in params. 
    This visualisation contains all the columns in the X and Y axis, where the 
    intersection of the column and row displays the correlation value. 
    The closer this correlation factor is to 1, the more correlated the features
    are. If you would like to visualise this as a tree, please see 
    the function _plot_correlation_dendogram.

    :Example:

    columns_to_include=["age", "loan", "gender"]
    xai._plot_correlation_matrix(df, cols=columns_to_include)

    :returns: Null
    :rtype: None

    """
    fig = plt.figure(**plt_kwargs)
    ax = fig.add_subplot(111)
    cax = ax.matshow(
            corr, 
            cmap='coolwarm', 
            vmin=-1, 
            vmax=1)
    fig.colorbar(cax)
    ticks = np.arange(0,len(cols),1)
    ax.set_xticks(ticks)
    plt.xticks(rotation=90)
    ax.set_yticks(ticks)
    ax.set_xticklabels(cols)
    ax.set_yticklabels(cols)
    plt.show()

def correlations(
        df: pd.DataFrame,
        include_categorical: bool = False,
        plot_type: str = "dendogram",
        plt_kwargs={},
        categorical_cols: List[str] = []):
    """
    Computes the correlations for the columns provided and plots the relevant
    image as requested by the parameters.

    :Example:

    cat_df = xai.balance(
        df, 
        "gender", "loan",
        upsample=0.8,
        downsample=0.8)

    :param df: Pandas Dataframe containing data (inputs and target )
    :type df: pandas.DataFrame
    :param *cross_cols: One or more positional arguments (passed as *args) that 
    are used to split the data into the cross product of their values 
    :type cross_cols: List[str]
    :param upsample: [Default: 0.5] Target upsample for columns lower 
        than percentage.
    :type upsample: float
    :param downsample: [Default: 1] Target downsample for columns higher 
        than percentage.
    :type downsample: float
    :param bins: [Default: 6] Number of bins to be used for numerical cols
    :type bins: int
    :param categorical_cols: [Default: []] Columns within dataframe that are
        categorical. Columns that are not np.objects and are not part explicitly
        provided here will be treated as numeric, and bins will be used.
    :type categorical_cols: List[str]
    :param threshold: [Default: 0.5] Threshold to display in the chart.
    :type threshold: float
    :returns: Returns a dataframe containing the correlation values for the features
    :rtype: pandas.DataFrame

    """

    corr = None
    cols: List = []
    if include_categorical:
        corr = sr(df).correlation 
        cols = df.columns
    else:

        if not len(categorical_cols):
            categorical_cols = df.select_dtypes(include=[np.object, np.bool]).columns

        cols = [c for c in df.columns if c not in categorical_cols]

        corr = df[cols].corr()
        cols = corr.columns

    if plot_type == "dendogram":
        _plot_correlation_dendogram(corr, cols, plt_kwargs=plt_kwargs)
    elif plot_type == "matrix":
        _plot_correlation_matrix(corr, cols, plt_kwargs=plt_kwargs)
    else:
        raise ValueError(f"Variable plot_type not valid. Provided: {plot_type}")

    return corr

def confusion_matrix_plot(
        y_test, 
        pred, 
        scaled=True,
        label_x_neg="PREDICTED NEGATIVE",
        label_x_pos="PREDICTED POSITIVE", 
        label_y_neg="ACTUAL NEGATIVE", 
        label_y_pos="ACTUAL POSITIVE"):
    """
    Plots a confusion matrix for a binary classifier with the expected and 
    predicted values provided.

    :Example:

    xai.confusion_matrix_plot(
        actual_labels, 
        predicted_labels,
        scaled=True)

    :param y_test: Array containing binary "actual" labels for data
    :type y_test: Union[np.array, list]
    :param pred: Array containing binary "predictedd" labels for data
    :type pred: Union[np.array, list]
    :param scaled: [Default: True] Whether the values are scaled to 0-1 or 
    displayed as total number of instances
    :type scaled: bool
    :param label_x_neg: [Default: "PREDICTED NEGATIVE"] Plot label for 
    the predicted negative values
    :type label_x_neg: str
    :param label_x_pos: [Default: "PREDICTED POSITIVE"] Plot label for 
    the predicted positive values
    :type label_x_pos: str
    :param label_y_neg: [Default: "ACTUAL NEGATIVE"] Plot label for 
    the actual negative values
    :type label_y_neg: str
    :param label_y_pos: [Default: "ACTUAL POSITIVE"] Plot label for 
    the actual positive values
    :type label_y_pos: str
    :returns: Null
    :rtype: None
    """

    confusion = confusion_matrix(y_test, pred)
    columns = [label_y_neg, label_y_pos]
    index = [label_x_neg, label_x_pos]

    if scaled:
        confusion_scaled = (confusion.astype("float") / 
                            confusion.sum(axis=1)[:, np.newaxis])
        confusion = pd.DataFrame(
                confusion_scaled, 
                index=index, 
                columns=columns)
    else:
        confusion = pd.DataFrame(
                confusion,
                index=index,
                columns=columns)

    cmap = plt.get_cmap("Blues")
    plt.figure()
    plt.imshow(confusion, interpolation="nearest", cmap=cmap)
    plt.title("Confusion matrix")
    plt.colorbar()

    plt.xticks(np.arange(2), columns, rotation=45)
    plt.yticks(np.arange(2), index, rotation=45)

    threshold = 0.5 if scaled else confusion.max().max() / 2
    for i, j in itertools.product(
            range(confusion.shape[0]), 
            range(confusion.shape[1])):
        txt = "{:,}".format(confusion.iloc[i,j])
        if scaled: txt = "{:0.4f}".format(confusion.iloc[i,j])
        plt.text(j, i, txt,
                    horizontalalignment="center",
                    color=("white" if confusion.iloc[i,j] > threshold else "black"))

    plt.tight_layout()
    plt.show()


def balanced_train_test_split(
        x: pd.DataFrame,
        y: Union[np.ndarray, list],
        *cross_cols: str,
        categorical_cols: List[str] = [],
        min_per_group: int = 20,
        max_per_group: Optional[int] = None,
        fallback_type: str = "upsample",
        bins: int =6, 
        random_state: int=None
        ) -> Tuple[
                pd.DataFrame, 
                np.ndarray, 
                pd.DataFrame, 
                np.ndarray, 
                np.ndarray,
                np.ndarray]:
    """
    Splits the "x" DataFrame and "y" Array into train/test split training sets with 
    a balanced number of examples for each of the categories of the columns provided.
    For example, if the columns provided are "gender" and "loan", the resulting splits
    would contain an equal number of examples for Male with Loan Approved, Male with 
    Loan Rejected, Female with Loan Approved, and Female with Loan Rejected. The 
    "fallback_type" parameter provides the behaviour that is triggered if there are not
    enough datapoint examples for one of the subcategory groups - the default is "half"

    Example
    -------

    .. code-block:: python

        x: pd.DataFrame # Contains the input features
        y: np.array # Contains the labels for the data
        categorical_cols: List[str] # Name of columns that are categorical

        x_train, y_train, x_test, y_test, train_idx, test_idx = \\
            xai.balanced_train_test_split(
                    x, y, balance_on=["gender"], 
                    categorical_cols=categorical_cols, min_per_group=300,
                    fallback_type="half")

    Args
    -----
    x : 
        Pandas dataframe containing all the features in dataset
    y : 
        Array containing "actual" labels for the dataset 
    *cross_cols :
        One or more positional arguments (passed as *args) that 
        are used to split the data into the cross product of their values 
    categorical_cols : 
        [Default: []] Columns within dataframe that are
        categorical. Columns that are not np.objects and are not part explicitly
        provided here will be treated as numeric, and bins will be used.
    min_per_group : 
        [Default: 20] This is the number of examples for each
        of the groups created
    max_per_group : 
        [Default: None] This is the maximum number of examples for
        each group to be provided with.
    fallback_type : 
        [Default: upsample] This is the fallback mechanism for when
        one of the groups contains less elements than the number provided 
        through min_per_group. The options are "upsample", "ignore" and "error". 
            - "upsample": This will get samples with replacement so will repeat elements
            - "ignore": Will just ignore and return all the elements available
            - "error": Throw an exception for any groups with less elements
    bins : 
        [Default: 6] Number of bins to be used for numerical cols
    random_state: 
        [Default: None] Random seed for the internal sampling

    Returns
    -------

    x_train : pd.DataFrame 
        DataFrame containing traning datapoints
    y_train : np.ndarray 
        Array containing labels for training datapoints
    x_test : pd.DataFrame
        DataFrame containing test datapoints
    y_test : np.ndarray
        Array containing labels for test datapoints
    train_idx : np.ndarray 
        Boolean array with True on Training indexes
    test_idx : np.ndarray 
        Boolean array with True on Testing indexes

    """

    if not cross_cols:
        raise TypeError("imbalance_plot requires at least 1 string column name")
    if min_per_group < 1:
        raise TypeError("min_per_group must be at least 1")
    if max_per_group and max_per_group < min_per_group:
        raise TypeError(f"min_per_group ({min_per_group}) must be less or equal than "
                f"max_per_group ({max_per_group}) if max_per_group is provided.")
    
    if random_state:
        random.setstate(random_state)

    tmp_df = x.copy()
    tmp_df["target"] = y

    cross = ["target"] + list(cross_cols)
    
    if not categorical_cols:
        categorical_cols = _infer_categorical(tmp_df)

    # TODO: Enable for non-categorical targets
    categorical_cols = ["target"] + categorical_cols

    grouped = group_by_columns(
            tmp_df, 
            cross, 
            bins=bins,
            categorical_cols=categorical_cols)

    def resample(x):
        group_size = x.shape[0]

        if max_per_group:
            if group_size > max_per_group:
                return x.sample(max_per_group)

        if group_size > min_per_group:
            return x.sample(min_per_group)

        if fallback_type == "upsample":
            return x.sample(min_per_group, replace=True)
        elif fallback_type == "ignore":
            return x
        elif fallback_type == "error":
            raise ValueError("Number of samples for group are not enough,"
                    " and fallback_type provided was 'error'")
        else:
            raise(f"Sampling type provided not found: given {fallback_type}, "\
                 "expected: 'error', or 'half'")
                    
    group = grouped.apply(resample)

    selected_idx = [g[-1] for g in group.index.values]
    
    train_idx = np.full(tmp_df.shape[0], True, dtype=bool)
    train_idx[selected_idx] = False
    test_idx = np.full(tmp_df.shape[0], False, dtype=bool)
    test_idx[selected_idx] = True
    
    df_train = tmp_df.iloc[train_idx] 
    df_test = tmp_df.iloc[test_idx]

    x_train = df_train.drop("target", axis=1)
    y_train = df_train["target"].values
    x_test = df_test.drop("target", axis=1)
    y_test = df_test["target"].values
    
    return x_train, y_train, x_test, y_test, train_idx, test_idx


def convert_probs(
        probs: np.ndarray,
        threshold: float = 0.5
        ) -> np.ndarray:
    """
    Converts all the probabilities in the array provided into binary labels
    as per the threshold provided which is 0.5 by default.

    Example
    ---------

    .. code-block:: python

        probs = np.array([0.1, 0.2, 0.7, 0.8, 0.6])
        labels = xai.convert_probs(probs, threshold=0.65)
        print(labels)

        > [0, 0, 1, 1, 0]
        
    Args
    -------

    probs : 
        Numpy array or list containing a list of floats between 0 and 1
    threshold :
        Float that provides the threshold for which probabilities over the 
        threshold will be converted to 1

    Returns
    ----------

    : np.ndarray
        Numpy array containing the labels based on threshold provided

    """
    
    return (probs >= threshold).astype(int)

def evaluation_metrics(
        y_valid, 
        y_pred
        ) -> Dict[str, float]:
    """
    Calculates model performance metrics (accuracy, precision, recall, etc) 
    from the actual and predicted lables provided. 

    Example
    ---------

    .. code-block:: python

        y_actual: np.ndarray
        y_predicted: np.ndarray

        metrics = xai.evaluation_metrics(y_actual, y_predicted)
        for k,v in metrics.items():
            print(f"{k}: {v}")

        > precision: 0.8, 
        > recall: 0.9, 
        > specificity: 0.7, 
        > accuracy: 0.8, 
        > auc: 0.7, 
        > f1: 0.8
        
    Args
    -------

    y_valid : 
        Numpy array with the actual labels for the datapoints
    y_pred :
        Numpy array with the predicted labels for the datapoints

    Returns
    ----------

    : Dict[str, float]
        Dictionary containing the metrics as follows:

        .. code-block:: python

            return {
                "precision": precision, 
                "recall": recall, 
                "specificity": specificity, 
                "accuracy": accuracy, 
                "auc": auc, 
                "f1": f1
            }

    """

    TP = np.sum( y_pred[y_valid==1] )
    TN = np.sum( y_pred[y_valid==0] == 0 )
    FP = np.sum(  y_pred[y_valid==0] )
    FN = np.sum(  y_pred[y_valid==1] == 0 )

    # Adding an OR to ensure it doesn't divide by zero
    precision = TP / ((TP+FP) or 0.001)
    recall = TP / ((TP+FN) or 0.001)
    specificity = TN / ((TN+FP) or 0.001)
    accuracy = (TP+TN) / (TP+TN+FP+FN)
    f1 = 2 * (precision * recall) / ((precision + recall) or 0.001)
    try:
        auc = roc_auc_score(y_valid, y_pred)
    except ValueError:
        auc = 0

    return {
        "precision": precision, 
        "recall": recall, 
        "specificity": specificity, 
        "accuracy": accuracy, 
        "auc": auc, 
        "f1": f1
    }

def metrics_plot(
        target: np.ndarray,
        predicted: np.ndarray,
        df: pd.DataFrame = pd.DataFrame(),
        cross_cols: List[str] = [],
        categorical_cols: List[str] = [],
        bins: int = 6,
        plot: bool = True,
        exclude_metrics: List[str] = [],
        plot_threshold: float = 0.5
        ) -> pd.DataFrame:
    """
    Creates a plot that displays statistical metrics including precision, 
    recall, accuracy, auc, f1 and specificity for each of the groups created
    for the columns provided by cross_cols. For example, if the columns passed
    are "gender" and "age", the resulting plot will show the statistical metrics
    for Male and Female for each binned group.

    Example
    ---------

    .. code-block:: python

        target: np.ndarray
        predicted: np.ndarray

        df_metrics = xai.metrics_plot(
                        target,
                        predicted,
                        df=df_data,
                        cross_cols=["gender", "age"],
                        bins=3

    Args
    -------

    target: 
        Numpy array containing the target labels for the datapoints
    predicted :
        Numpy array containing the predicted labels for the datapoints
    df :
        Pandas dataframe containing all the features for the datapoints.
        It can be empty if only looking to calculate global metrics, but
        if you would like to compute for categories across columns, the
        columns you are grouping by need to be provided
    cross_cols :
        Contains the columns that you would like to use to cross the values
    bins :
        [Default: 6] The number of bins in which you'd like 
        numerical columns to be split
    plot :
        [Default: True] If True a plot will be drawn with the results
    exclude_metrics :
        These are the metrics that you can choose to exclude if you only
        want specific ones (for example, excluding "f1", "specificity", etc)
    plot_threshold:
        The percentage that will be used to draw the threshold line in the plot
        which would provide guidance on what is the ideal metrics to achieve.

    Returns
    ----------

    : pd.DataFrame
        Pandas Dataframe containing all the metrics for the groups provided

    """

    grouped = _group_metrics(
        target,
        predicted,
        df, 
        cross_cols, 
        categorical_cols,
        bins,
        target_threshold=plot_threshold)

    prfs = []
    classes = []
    for group, group_df in grouped:
        group_valid = group_df['target'].values
        group_pred = group_df["predicted"].values
        metrics_dict = \
            evaluation_metrics(group_valid, group_pred)
        # Remove metrics as specified by params
        [metrics_dict.pop(k, None) for k in exclude_metrics]
        prfs.append(list(metrics_dict.values()))
        classes.append(str(group))

    prfs_cols = metrics_dict.keys()
    prfs_df = pd.DataFrame(
            np.array(prfs).transpose(), 
            columns=classes, 
            index=prfs_cols)

    if plot:
        prfs_df.plot.bar(figsize=(20,5))
        lp = plt.axhline(0.5, color='r')
        lp = plt.axhline(1, color='g')

    return prfs_df

def roc_plot(
        target,
        predicted,
        df=pd.DataFrame(),
        cross_cols=[],
        categorical_cols=[],
        bins=6,
        plot=True):

    return _curve(
        target=target,
        predicted=predicted,
        curve_type="roc",
        df=df,
        cross_cols=cross_cols,
        categorical_cols=categorical_cols,
        bins=bins,
        plot=plot)

def pr_plot(
        target,
        predicted,
        df=pd.DataFrame(),
        cross_cols=[],
        categorical_cols=[],
        bins=6,
        plot=True):

    return _curve(
        target=target,
        predicted=predicted,
        curve_type="pr",
        df=df,
        cross_cols=cross_cols,
        categorical_cols=categorical_cols,
        bins=bins,
        plot=plot)

def _curve(
        target,
        predicted,
        curve_type="roc",
        df=pd.DataFrame(),
        cross_cols=[],
        categorical_cols=[],
        bins=6,
        plot=True):

    if curve_type == "roc":
        curve_func = roc_curve
        y_label = 'False Positive Rate'
        x_label = 'True Positive Rate'
        p1 = [0,1]
        p2 = [0,1]
        y_lim = [0, 1.05]
        legend_loc = "lower right"
    elif curve_type == "pr":
        curve_func = precision_recall_curve 
        y_label = "Recall"
        x_label = "Precision"
        p1 = [1,0]
        p2 = [0.5,0.5]
        y_lim = [0.25, 1.05]
        legend_loc = "lower left"
    else:
        raise ValueError("Curve function provided not valid. "
                f" curve_func provided: {curve_func}")

    grouped = _group_metrics(
        target,
        predicted,
        df, 
        cross_cols, 
        categorical_cols,
        bins)

    if plot:
        plt.figure()

    r1s = r2s = []

    for group, group_df in grouped:
        group_valid = group_df["target"]
        group_pred = group_df["predicted"]

        r1, r2, _ = curve_func(group_valid, group_pred)
        r1s.append(r1)
        r2s.append(r2)

        if plot:
            if curve_type == "pr": r1,r2 = r2,r1
            plt.plot(r1, r2, label=group)
            plt.plot(p1, p2, 'k--')

    if plot:
        plt.xlim([0.0, 1.0])
        plt.ylim(y_lim)
        plt.xlabel(x_label)
        plt.ylabel(y_label)
        plt.legend(loc=legend_loc)
        plt.show()

    return r1s, r2s


def _infer_categorical(df):
    categorical_cols = df.select_dtypes(
            include=[np.object, np.bool, np.int8]).columns
    logging.warn("No categorical_cols passed so inferred using np.object, "
            f"np.int8 and np.bool: {categorical_cols}. If you see an error"
            " these are not "
            "correct, please provide them as a string array as: "
            "categorical_cols=['col1', 'col2', ...]")
    return categorical_cols


def _group_metrics(
        target,
        predicted,
        df, 
        cross_cols, 
        categorical_cols,
        bins,
        target_threshold=None):

    if not all(c in df.columns for c in cross_cols):
        raise KeyError("Cross columns don't match columns in dataframe provided.")

    df_tmp = df.copy()
    df_tmp["target"] = target
    df_tmp["predicted"] = predicted

    # Convert predictions into classes
    if target_threshold and df_tmp["predicted"].dtype.kind == 'f':
        df_tmp["predicted"] = convert_probs(
            df_tmp["predicted"], threshold=target_threshold)

    if not categorical_cols and cross_cols:
        categorical_cols = _infer_categorical(df_tmp)

    if not cross_cols:
        grouped = [("target", df_tmp),]
    else:
        grouped = group_by_columns(
            df_tmp,
            cross_cols,
            bins=bins,
            categorical_cols=categorical_cols)

    return grouped



def smile_imbalance(
        y_test, 
        probs: np.ndarray, 
        threshold=0.5, 
        manual_review=None,
        display_breakdown=False,
        bins=10):
    
    # TODO: Change function so it only iterates once
    probs = np.array(probs)
    preds = convert_probs(probs, threshold).flatten()
    d = pd.DataFrame(probs)
    d.columns = ["probs"]
    d["preds"] = preds
    d["target"] = y_test

    tps = np.full(y_test.shape, False, bool)

    d["true-positives"] = np.full(y_test.shape[0], False, bool)
    d["true-negatives"] = np.full(y_test.shape[0], False, bool)
    d["false-positives"] = np.full(y_test.shape[0], False, bool)
    d["false-negatives"] = np.full(y_test.shape[0], False, bool)
    d["manual-review"] =  np.full(y_test.shape[0], False, bool)

    d["true-positives"].loc[y_test == 1] = preds[y_test == 1] == 1
    d["true-negatives"].loc[y_test == 0] = preds[y_test == 0] == 0
    d["false-positives"].loc[y_test == 0] = preds[y_test == 0] == 1
    d["false-negatives"].loc[y_test == 1] = preds[y_test == 1] == 0

    d["correct"] = d["true-positives"].values
    d["correct"].loc[d["true-negatives"] == 1] = True

    d["incorrect"] = d["false-positives"].values
    d["incorrect"].loc[d["false-negatives"] == 1] = True 
    
    if display_breakdown:
        disp_cols = ["true-positives", 
                     "true-negatives", 
                     "false-positives", 
                     "false-negatives"]
    else:
        disp_cols = ["correct", "incorrect"]
    
    if manual_review:
        gt = probs > manual_review
        lt = probs < threshold
        d["manual-review"] = gt * lt > 0
        
        if display_breakdown:
            d["true-positives"].loc[d["manual-review"]] = False
            d["true-negatives"].loc[d["manual-review"]] = False
            d["false-positives"].loc[d["manual-review"]] = False
            d["false-negatives"].loc[d["manual-review"]] = False
        else:
            d["correct"].loc[d["manual-review"]] = False
            d["incorrect"].loc[d["manual-review"]] = False
        
        disp_cols.append("manual-review")

    d["true-positives"] = d["true-positives"].astype(int) 
    d["true-negatives"] = d["true-negatives"].astype(int)
    d["false-positives"] = d["false-positives"].astype(int)
    d["false-negatives"] = d["false-negatives"].astype(int)
    d["correct"] = d["correct"].astype(int)
    d["incorrect"] = d["incorrect"].astype(int)

    grouped = group_by_columns(d, ["probs"], bins=bins)

    ax = grouped[disp_cols].sum().plot.bar(stacked=True, figsize=(15,5))
    lim = ax.get_xlim()
    ran = lim[1] - lim[0]
    thre = ran*threshold + lim[0]
    plt.axvline(thre)
    if manual_review:
        manr = ran*manual_review + lim[0]
        plt.axvline(manr)
    # TODO: Need to fix this hack and use the index
    ax_xticks = [label.get_text().split()[1][:-1] for label in ax.get_xticklabels()]
    ax.set_xticklabels(ax_xticks)
    
    return d


def feature_importance(x, y, func, repeat=10, plot=True):
    base_score = func(x, y)
    imp = [0] * len(x.columns)
    for i in range(repeat):
        for j, c in enumerate(x.columns):
            tmp = x[c].values.copy()
            np.random.shuffle(x[c].values)
            score = func(x, y)
            x[c] = tmp
            imp[j] += base_score - score
    imp = [a/repeat for a in imp]
    imp_df = pd.DataFrame(data=[imp], columns=x.columns)
    if plot:
        imp_df.sum().sort_values().plot.barh()
    return imp_df