klib.utils._corr_selector() - Code Metrics - Inspection of "set default values for high and low correlations" - akanz1/klib - Measure and Improve Code Quality continuously with Scrutinizer

Passed

Push — master ( 32c39e...8bebb3 )

by Andreas

created 2020-08-02 13:16 UTC

klib.utils._corr_selector() B

↳ Parent: klib.utils

Complexity

Conditions

Size

Total Lines	39
Code Lines	19

Duplication

Lines	0
Ratio	0 %

Importance

Changes

Metric	Value
cc	7
eloc	19
nop	3
dl	0
loc	39
rs	8
c	0
b	0
f	0

"""
Utilities and auxiliary functions.

:author: Andreas Kanz

"""

# Imports
import numpy as np
import pandas as pd
from typing import Any, Dict, List, Optional, Tuple, Union


def _corr_selector(
    corr: Union[pd.Series, pd.DataFrame],
    split: Optional[str] = None,  # Optional[Literal["pos", "neg", "above", "below"]] = None,
    threshold: float = 0,
) -> Union[pd.Series, pd.DataFrame]:
    """[summary]

    Parameters
    ----------
    corr : Union[pd.Series, pd.DataFrame]
        pd.Series or pd.DataFrame of correlations
    split : Optional[str], optional
        Type of split performed, by default None
           * {None, 'pos', 'neg', 'high', 'low'}
    threshold : float, optional
        Value between 0 and 1 to set the correlation threshold, by default 0

    Returns
    -------
    pd.DataFrame
        List or matrix of (filtered) correlations
    """

    if split == "pos":
        corr = corr.where((corr >= threshold) & (corr > 0))
        print('Displaying positive correlations. Use a positive "threshold" to further limit the results.')
    elif split == "neg":
        corr = corr.where((corr <= threshold) & (corr < 0))
        print('Displaying negative correlations. Use a negative "threshold" to further limit the results.')
    elif split == "high":
        threshold = 0.5 if threshold == 0 else threshold
        corr = corr.where(np.abs(corr) >= threshold)
        print(f"Displaying absolute correlations above the threshold ({threshold}).")
    elif split == "low":
        threshold = 0.5 if threshold == 0 else threshold
        corr = corr.where(np.abs(corr) <= threshold)
        print(f"Displaying absolute correlations below the threshold ({threshold}).")

    return corr


def _diff_report(
    data: pd.DataFrame,
    data_cleaned: pd.DataFrame,
    dupl_rows: Optional[List[Union[str, int]]] = None,
    single_val_cols: Optional[List[str]] = None,
    show: Optional[str] = "changes",  # Optional[Literal["all", "changes"]] = "changes",
) -> None:
    """ Provides information about changes between two datasets, such as dropped rows and columns, memory usage and \
    missing values.

    Parameters
    ----------
    data : pd.DataFrame
        2D dataset that can be coerced into Pandas DataFrame. Input the initial dataset here
    data_cleaned : pd.DataFrame
        2D dataset that can be coerced into Pandas DataFrame. Input the cleaned / updated dataset here
    dupl_rows : Optional[List[Union[str, int]]], optional
        List of duplicate row indices, by default None
    single_val_cols : Optional[List[str]], optional
        List of single-valued column indices. I.e. columns where all cells contain the same value. \
        NaNs count as a separate value, by default None
    show : str, optional
        {'all', 'changes', None}, by default "changes"
        Specify verbosity of the output:
            * 'all': Print information about the data before and after cleaning as well as information about changes \
            and memory usage (deep). Please be aware, that this can slow down the function by quite a bit.
            * 'changes': Print out differences in the data before and after cleaning.
            * None: No information about the data and the data cleaning is printed.

    Returns
    -------
    None
        Print statement highlighting the datasets or changes between the two datasets.
    """

    if show in ["changes", "all"]:
        dupl_rows = [] if dupl_rows is None else dupl_rows.copy()
        single_val_cols = [] if single_val_cols is None else single_val_cols.copy()
        data_mem = _memory_usage(data, deep=False)
        data_cl_mem = _memory_usage(data_cleaned, deep=False)
        data_mv_tot = _missing_vals(data)["mv_total"]
        data_cl_mv_tot = _missing_vals(data_cleaned)["mv_total"]

        if show == "all":
            data_mem = _memory_usage(data, deep=True)
            data_cl_mem = _memory_usage(data_cleaned, deep=True)
            print("Before data cleaning:\n")
            print(f"dtypes:\n{data.dtypes.value_counts()}")
            print(f"\nNumber of rows: {data.shape[0]}")
            print(f"Number of cols: {data.shape[1]}")
            print(f"Missing values: {data_mv_tot}")
            print(f"Memory usage: {data_mem} MB")
            print("_______________________________________________________\n")
            print("After data cleaning:\n")
            print(f"dtypes:\n{data_cleaned.dtypes.value_counts()}")
            print(f"\nNumber of rows: {data_cleaned.shape[0]}")
            print(f"Number of cols: {data_cleaned.shape[1]}")
            print(f"Missing values: {data_cl_mv_tot}")
            print(f"Memory usage: {data_cl_mem} MB")
            print("_______________________________________________________\n")

        print(f"Shape of cleaned data: {data_cleaned.shape} - Remaining NAs: {data_cl_mv_tot}")
        print("\nChanges:")
        print(f"Dropped rows: {data.shape[0]-data_cleaned.shape[0]}")
        print(f"     of which {len(dupl_rows)} duplicates. (Rows: {dupl_rows})")
        print(f"Dropped columns: {data.shape[1]-data_cleaned.shape[1]}")
        print(f"     of which {len(single_val_cols)} single valued. (Columns: {single_val_cols})")
        print(f"Dropped missing values: {data_mv_tot-data_cl_mv_tot}")
        mem_change = data_mem - data_cl_mem
        print(f"Reduced memory by at least: {round(mem_change,3)} MB (-{round(100*mem_change/data_mem,2)}%)")


def _drop_duplicates(data: pd.DataFrame) -> Tuple[pd.DataFrame, Any]:
    """ Provides information on and drops duplicate rows.

    Parameters
    ----------
    data : pd.DataFrame
        2D dataset that can be coerced into Pandas DataFrame

    Returns
    -------
    Tuple[pd.DataFrame, List]
        Deduplicated Pandas DataFrame and Index Object of rows dropped
    """

    data = pd.DataFrame(data).copy()
    dupl_rows = data[data.duplicated()].index.tolist()
    data = data.drop(dupl_rows, axis="index")

    return data, dupl_rows


def _memory_usage(data: pd.DataFrame, deep: bool = True) -> float:
    """ Gives the total memory usage in megabytes.

    Parameters
    ----------
    data : pd.DataFrame
        2D dataset that can be coerced into Pandas DataFrame
    deep : bool, optional
        Runs a deep analysis of the memory usage, by default True

    Returns
    -------
    float
        Memory usage in megabytes
    """

    data = pd.DataFrame(data).copy()
    memory_usage = round(data.memory_usage(index=True, deep=deep).sum() / (1024 ** 2), 2)

    return memory_usage


def _missing_vals(data: pd.DataFrame) -> Dict[str, Any]:
    """ Gives metrics of missing values in the dataset.

    Parameters
    ----------
    data : pd.DataFrame
        2D dataset that can be coerced into Pandas DataFrame

    Returns
    -------
    Dict[str, float]
        mv_total: float, number of missing values in the entire dataset
        mv_rows: float, number of missing values in each row
        mv_cols: float, number of missing values in each column
        mv_rows_ratio: float, ratio of missing values for each row
        mv_cols_ratio: float, ratio of missing values for each column
    """

    data = pd.DataFrame(data).copy()
    mv_rows = data.isna().sum(axis=1)
    mv_cols = data.isna().sum(axis=0)
    mv_total = data.isna().sum().sum()
    mv_rows_ratio = mv_rows / data.shape[1]
    mv_cols_ratio = mv_cols / data.shape[0]

    return {
        "mv_total": mv_total,
        "mv_rows": mv_rows,
        "mv_cols": mv_cols,
        "mv_rows_ratio": mv_rows_ratio,
        "mv_cols_ratio": mv_cols_ratio,
    }


def _validate_input_bool(value, desc):
    if not (isinstance(value, bool)):
        raise TypeError(f"Input value for '{desc}' is {type(value)} but should be a boolean.")


def _validate_input_int(value, desc):
    if not isinstance(value, int):
        raise TypeError(f"Input value for '{desc}' is {type(value)} but should be an integer.")


def _validate_input_range(value, desc, lower, upper):
    if value < lower or value > upper:
        raise ValueError(f"'{desc}' = {value} but should be within the range {lower} <= '{desc}' <= {upper}.")


def _validate_input_smaller(value1, value2, desc):
    if value1 > value2:
        raise ValueError(f"The first input for '{desc}' should be smaller or equal to the second input.")


def _validate_input_sum(limit, desc, *args):
    if sum(args) > limit:
        raise ValueError(f"The sum of imput values provided for '{desc}' should be less or equal to {limit}.")


1			"""
2			Utilities and auxiliary functions.
3
4			:author: Andreas Kanz
5
6			"""
7
8			# Imports
9			import numpy as np
10			import pandas as pd
11			from typing import Any, Dict, List, Optional, Tuple, Union
12
13
14			def _corr_selector(
15			corr: Union[pd.Series, pd.DataFrame],
16			split: Optional[str] = None, # Optional[Literal["pos", "neg", "above", "below"]] = None,
17			threshold: float = 0,
18			) -> Union[pd.Series, pd.DataFrame]:
19			"""[summary]
20
21			Parameters
22			----------
23			corr : Union[pd.Series, pd.DataFrame]
24			pd.Series or pd.DataFrame of correlations
25			split : Optional[str], optional
26			Type of split performed, by default None
27			* {None, 'pos', 'neg', 'high', 'low'}
28			threshold : float, optional
29			Value between 0 and 1 to set the correlation threshold, by default 0
30
31			Returns
32			-------
33			pd.DataFrame
34			List or matrix of (filtered) correlations
35			"""
36
37			if split == "pos":
38			corr = corr.where((corr >= threshold) & (corr > 0))
39			print('Displaying positive correlations. Use a positive "threshold" to further limit the results.')
40			elif split == "neg":
41			corr = corr.where((corr <= threshold) & (corr < 0))
42			print('Displaying negative correlations. Use a negative "threshold" to further limit the results.')
43			elif split == "high":
44			threshold = 0.5 if threshold == 0 else threshold
45			corr = corr.where(np.abs(corr) >= threshold)
46			print(f"Displaying absolute correlations above the threshold ({threshold}).")
47			elif split == "low":
48			threshold = 0.5 if threshold == 0 else threshold
49			corr = corr.where(np.abs(corr) <= threshold)
50			print(f"Displaying absolute correlations below the threshold ({threshold}).")
51
52			return corr
53
54
55			def _diff_report(
56			data: pd.DataFrame,
57			data_cleaned: pd.DataFrame,
58			dupl_rows: Optional[List[Union[str, int]]] = None,
59			single_val_cols: Optional[List[str]] = None,
60			show: Optional[str] = "changes", # Optional[Literal["all", "changes"]] = "changes",
61			) -> None:
62			""" Provides information about changes between two datasets, such as dropped rows and columns, memory usage and \
63			missing values.
64
65			Parameters
66			----------
67			data : pd.DataFrame
68			2D dataset that can be coerced into Pandas DataFrame. Input the initial dataset here
69			data_cleaned : pd.DataFrame
70			2D dataset that can be coerced into Pandas DataFrame. Input the cleaned / updated dataset here
71			dupl_rows : Optional[List[Union[str, int]]], optional
72			List of duplicate row indices, by default None
73			single_val_cols : Optional[List[str]], optional
74			List of single-valued column indices. I.e. columns where all cells contain the same value. \
75			NaNs count as a separate value, by default None
76			show : str, optional
77			{'all', 'changes', None}, by default "changes"
78			Specify verbosity of the output:
79			* 'all': Print information about the data before and after cleaning as well as information about changes \
80			and memory usage (deep). Please be aware, that this can slow down the function by quite a bit.
81			* 'changes': Print out differences in the data before and after cleaning.
82			* None: No information about the data and the data cleaning is printed.
83
84			Returns
85			-------
86			None
87			Print statement highlighting the datasets or changes between the two datasets.
88			"""
89
90			if show in ["changes", "all"]:
91			dupl_rows = [] if dupl_rows is None else dupl_rows.copy()
92			single_val_cols = [] if single_val_cols is None else single_val_cols.copy()
93			data_mem = _memory_usage(data, deep=False)
94			data_cl_mem = _memory_usage(data_cleaned, deep=False)
95			data_mv_tot = _missing_vals(data)["mv_total"]
96			data_cl_mv_tot = _missing_vals(data_cleaned)["mv_total"]
97
98			if show == "all":
99			data_mem = _memory_usage(data, deep=True)
100			data_cl_mem = _memory_usage(data_cleaned, deep=True)
101			print("Before data cleaning:\n")
102			print(f"dtypes:\n{data.dtypes.value_counts()}")
103			print(f"\nNumber of rows: {data.shape[0]}")
104			print(f"Number of cols: {data.shape[1]}")
105			print(f"Missing values: {data_mv_tot}")
106			print(f"Memory usage: {data_mem} MB")
107			print("_______________________________________________________\n")
108			print("After data cleaning:\n")
109			print(f"dtypes:\n{data_cleaned.dtypes.value_counts()}")
110			print(f"\nNumber of rows: {data_cleaned.shape[0]}")
111			print(f"Number of cols: {data_cleaned.shape[1]}")
112			print(f"Missing values: {data_cl_mv_tot}")
113			print(f"Memory usage: {data_cl_mem} MB")
114			print("_______________________________________________________\n")
115
116			print(f"Shape of cleaned data: {data_cleaned.shape} - Remaining NAs: {data_cl_mv_tot}")
117			print("\nChanges:")
118			print(f"Dropped rows: {data.shape[0]-data_cleaned.shape[0]}")
119			print(f" of which {len(dupl_rows)} duplicates. (Rows: {dupl_rows})")
120			print(f"Dropped columns: {data.shape[1]-data_cleaned.shape[1]}")
121			print(f" of which {len(single_val_cols)} single valued. (Columns: {single_val_cols})")
122			print(f"Dropped missing values: {data_mv_tot-data_cl_mv_tot}")
123			mem_change = data_mem - data_cl_mem
124			print(f"Reduced memory by at least: {round(mem_change,3)} MB (-{round(100*mem_change/data_mem,2)}%)")
125
126
127			def _drop_duplicates(data: pd.DataFrame) -> Tuple[pd.DataFrame, Any]:
128			""" Provides information on and drops duplicate rows.
129
130			Parameters
131			----------
132			data : pd.DataFrame
133			2D dataset that can be coerced into Pandas DataFrame
134
135			Returns
136			-------
137			Tuple[pd.DataFrame, List]
138			Deduplicated Pandas DataFrame and Index Object of rows dropped
139			"""
140
141			data = pd.DataFrame(data).copy()
142			dupl_rows = data[data.duplicated()].index.tolist()
143			data = data.drop(dupl_rows, axis="index")
144
145			return data, dupl_rows
146
147
148			def _memory_usage(data: pd.DataFrame, deep: bool = True) -> float:
149			""" Gives the total memory usage in megabytes.
150
151			Parameters
152			----------
153			data : pd.DataFrame
154			2D dataset that can be coerced into Pandas DataFrame
155			deep : bool, optional
156			Runs a deep analysis of the memory usage, by default True
157
158			Returns
159			-------
160			float
161			Memory usage in megabytes
162			"""
163
164			data = pd.DataFrame(data).copy()
165			memory_usage = round(data.memory_usage(index=True, deep=deep).sum() / (1024 ** 2), 2)
166
167			return memory_usage
168
169
170			def _missing_vals(data: pd.DataFrame) -> Dict[str, Any]:
171			""" Gives metrics of missing values in the dataset.
172
173			Parameters
174			----------
175			data : pd.DataFrame
176			2D dataset that can be coerced into Pandas DataFrame
177
178			Returns
179			-------
180			Dict[str, float]
181			mv_total: float, number of missing values in the entire dataset
182			mv_rows: float, number of missing values in each row
183			mv_cols: float, number of missing values in each column
184			mv_rows_ratio: float, ratio of missing values for each row
185			mv_cols_ratio: float, ratio of missing values for each column
186			"""
187
188			data = pd.DataFrame(data).copy()
189			mv_rows = data.isna().sum(axis=1)
190			mv_cols = data.isna().sum(axis=0)
191			mv_total = data.isna().sum().sum()
192			mv_rows_ratio = mv_rows / data.shape[1]
193			mv_cols_ratio = mv_cols / data.shape[0]
194
195			return {
196			"mv_total": mv_total,
197			"mv_rows": mv_rows,
198			"mv_cols": mv_cols,
199			"mv_rows_ratio": mv_rows_ratio,
200			"mv_cols_ratio": mv_cols_ratio,
201			}
202
203
204			def _validate_input_bool(value, desc):
205			if not (isinstance(value, bool)):
206			raise TypeError(f"Input value for '{desc}' is {type(value)} but should be a boolean.")
207
208
209			def _validate_input_int(value, desc):
210			if not isinstance(value, int):
211			raise TypeError(f"Input value for '{desc}' is {type(value)} but should be an integer.")
212
213
214			def _validate_input_range(value, desc, lower, upper):
215			if value < lower or value > upper:
216			raise ValueError(f"'{desc}' = {value} but should be within the range {lower} <= '{desc}' <= {upper}.")
217
218
219			def _validate_input_smaller(value1, value2, desc):
220			if value1 > value2:
221			raise ValueError(f"The first input for '{desc}' should be smaller or equal to the second input.")
222
223
224			def _validate_input_sum(limit, desc, *args):
225			if sum(args) > limit:
226			raise ValueError(f"The sum of imput values provided for '{desc}' should be less or equal to {limit}.")
227

akanz1 / klib

GitHub Access Token became invalid

Push — master ( 32c39e...8bebb3 )

klib.utils._corr_selector() B

Complexity

Size

Duplication

Importance

Duplication Side-by-Side

Filter issues like