Taiwanese Bankruptcy Prediction 🇹🇼

In [ ]:
__author__ = "Donald Ghazi"
__email__ = "donald@donaldghazi.com"
__website__ = "donaldghazi.com"
In [61]:
import gzip
import json
import pickle
import ipywidgets as widgets
import pandas as pd
import matplotlib.pyplot as plt

from imblearn.over_sampling import RandomOverSampler
from ipywidgets import interact
from sklearn.ensemble import GradientBoostingClassifier
from sklearn.impute import SimpleImputer
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import (
    ConfusionMatrixDisplay,
    classification_report,
    confusion_matrix,
)
from sklearn.model_selection import GridSearchCV, train_test_split, cross_val_score
from sklearn.pipeline import make_pipeline

Prepare Data¶

Import¶

I'll load the contents of the "data/taiwan-bankruptcy-data.json.gz" and assign it to the variable taiwan_data.

taiwan_data should be a dictionary. I'll create a DataFrame in this project.

In [62]:
%%bash

cd data
gzip -dkf taiwan-bankruptcy-data.json.gz
In [63]:
# Open file and load JSON
with gzip.open("data/taiwan-bankruptcy-data.json.gz", "r") as read_file:
    taiwan_data = json.load(read_file)
print(type(taiwan_data))

<class 'dict'>
In [64]:
# Extract the key names from taiwan_data and assign them to the variable taiwan_data_keys
taiwan_data_keys = taiwan_data.keys()
print(taiwan_data_keys)

dict_keys(['schema', 'metadata', 'observations'])

Now I want to calculate how many companies are in taiwan_data and assign the result to n_companies.

In [65]:
taiwan_data.keys()
Out[65]:
dict_keys(['schema', 'metadata', 'observations'])
In [66]:
type(taiwan_data["observations"])
Out[66]:
list
In [67]:
taiwan_data["observations"][0]
Out[67]:
{'id': 1,
 'bankrupt': True,
 'feat_1': 0.3705942573,
 'feat_2': 0.4243894461,
 'feat_3': 0.4057497725,
 'feat_4': 0.6014572133,
 'feat_5': 0.6014572133,
 'feat_6': 0.9989692032,
 'feat_7': 0.7968871459,
 'feat_8': 0.8088093609,
 'feat_9': 0.3026464339,
 'feat_10': 0.7809848502,
 'feat_11': 0.0001256969,
 'feat_12': 0.0,
 'feat_13': 0.4581431435,
 'feat_14': 0.0007250725,
 'feat_15': 0.0,
 'feat_16': 0.1479499389,
 'feat_17': 0.1479499389,
 'feat_18': 0.1479499389,
 'feat_19': 0.1691405881,
 'feat_20': 0.3116644267,
 'feat_21': 0.0175597804,
 'feat_22': 0.0959205276,
 'feat_23': 0.1387361603,
 'feat_24': 0.0221022784,
 'feat_25': 0.8481949945,
 'feat_26': 0.6889794628,
 'feat_27': 0.6889794628,
 'feat_28': 0.2175353862,
 'feat_29': 4980000000.0,
 'feat_30': 0.0003269773,
 'feat_31': 0.2630999837,
 'feat_32': 0.363725271,
 'feat_33': 0.0022589633,
 'feat_34': 0.0012077551,
 'feat_35': 0.629951302,
 'feat_36': 0.0212659244,
 'feat_37': 0.2075762615,
 'feat_38': 0.7924237385,
 'feat_39': 0.0050244547,
 'feat_40': 0.3902843544,
 'feat_41': 0.0064785025,
 'feat_42': 0.095884834,
 'feat_43': 0.1377573335,
 'feat_44': 0.3980356983,
 'feat_45': 0.0869565217,
 'feat_46': 0.0018138841,
 'feat_47': 0.0034873643,
 'feat_48': 0.0001820926,
 'feat_49': 0.0001165007,
 'feat_50': 0.0329032258,
 'feat_51': 0.034164182,
 'feat_52': 0.3929128695,
 'feat_53': 0.0371353016,
 'feat_54': 0.6727752925,
 'feat_55': 0.1666729588,
 'feat_56': 0.1906429591,
 'feat_57': 0.004094406,
 'feat_58': 0.0019967709,
 'feat_59': 0.000147336,
 'feat_60': 0.1473084504,
 'feat_61': 0.3340151713,
 'feat_62': 0.2769201582,
 'feat_63': 0.00103599,
 'feat_64': 0.6762691762,
 'feat_65': 0.7212745515,
 'feat_66': 0.3390770068,
 'feat_67': 0.025592368,
 'feat_68': 0.9032247712,
 'feat_69': 0.002021613,
 'feat_70': 0.0648557077,
 'feat_71': 701000000.0,
 'feat_72': 6550000000.0,
 'feat_73': 0.593830504,
 'feat_74': 458000000.0,
 'feat_75': 0.6715676536,
 'feat_76': 0.4242057622,
 'feat_77': 0.6762691762,
 'feat_78': 0.3390770068,
 'feat_79': 0.1265494878,
 'feat_80': 0.6375553953,
 'feat_81': 0.4586091477,
 'feat_82': 0.5203819179,
 'feat_83': 0.3129049481,
 'feat_84': 0.1182504766,
 'feat_85': 0,
 'feat_86': 0.7168453432,
 'feat_87': 0.00921944,
 'feat_88': 0.6228789594,
 'feat_89': 0.6014532901,
 'feat_90': 0.827890214,
 'feat_91': 0.2902018928,
 'feat_92': 0.0266006308,
 'feat_93': 0.5640501123,
 'feat_94': 1,
 'feat_95': 0.0164687409}
In [68]:
len(taiwan_data["observations"])
Out[68]:
6137
In [69]:
# Calculate how many companies are in taiwan_data and assign the result to n_companies
n_companies = len(taiwan_data["observations"])
print(n_companies)

6137

Now, I want to calculate the number of features associated with each company and assign the result to n_features.

In [70]:
type(taiwan_data["observations"])
Out[70]:
list
In [71]:
taiwan_data["observations"][0]
Out[71]:
{'id': 1,
 'bankrupt': True,
 'feat_1': 0.3705942573,
 'feat_2': 0.4243894461,
 'feat_3': 0.4057497725,
 'feat_4': 0.6014572133,
 'feat_5': 0.6014572133,
 'feat_6': 0.9989692032,
 'feat_7': 0.7968871459,
 'feat_8': 0.8088093609,
 'feat_9': 0.3026464339,
 'feat_10': 0.7809848502,
 'feat_11': 0.0001256969,
 'feat_12': 0.0,
 'feat_13': 0.4581431435,
 'feat_14': 0.0007250725,
 'feat_15': 0.0,
 'feat_16': 0.1479499389,
 'feat_17': 0.1479499389,
 'feat_18': 0.1479499389,
 'feat_19': 0.1691405881,
 'feat_20': 0.3116644267,
 'feat_21': 0.0175597804,
 'feat_22': 0.0959205276,
 'feat_23': 0.1387361603,
 'feat_24': 0.0221022784,
 'feat_25': 0.8481949945,
 'feat_26': 0.6889794628,
 'feat_27': 0.6889794628,
 'feat_28': 0.2175353862,
 'feat_29': 4980000000.0,
 'feat_30': 0.0003269773,
 'feat_31': 0.2630999837,
 'feat_32': 0.363725271,
 'feat_33': 0.0022589633,
 'feat_34': 0.0012077551,
 'feat_35': 0.629951302,
 'feat_36': 0.0212659244,
 'feat_37': 0.2075762615,
 'feat_38': 0.7924237385,
 'feat_39': 0.0050244547,
 'feat_40': 0.3902843544,
 'feat_41': 0.0064785025,
 'feat_42': 0.095884834,
 'feat_43': 0.1377573335,
 'feat_44': 0.3980356983,
 'feat_45': 0.0869565217,
 'feat_46': 0.0018138841,
 'feat_47': 0.0034873643,
 'feat_48': 0.0001820926,
 'feat_49': 0.0001165007,
 'feat_50': 0.0329032258,
 'feat_51': 0.034164182,
 'feat_52': 0.3929128695,
 'feat_53': 0.0371353016,
 'feat_54': 0.6727752925,
 'feat_55': 0.1666729588,
 'feat_56': 0.1906429591,
 'feat_57': 0.004094406,
 'feat_58': 0.0019967709,
 'feat_59': 0.000147336,
 'feat_60': 0.1473084504,
 'feat_61': 0.3340151713,
 'feat_62': 0.2769201582,
 'feat_63': 0.00103599,
 'feat_64': 0.6762691762,
 'feat_65': 0.7212745515,
 'feat_66': 0.3390770068,
 'feat_67': 0.025592368,
 'feat_68': 0.9032247712,
 'feat_69': 0.002021613,
 'feat_70': 0.0648557077,
 'feat_71': 701000000.0,
 'feat_72': 6550000000.0,
 'feat_73': 0.593830504,
 'feat_74': 458000000.0,
 'feat_75': 0.6715676536,
 'feat_76': 0.4242057622,
 'feat_77': 0.6762691762,
 'feat_78': 0.3390770068,
 'feat_79': 0.1265494878,
 'feat_80': 0.6375553953,
 'feat_81': 0.4586091477,
 'feat_82': 0.5203819179,
 'feat_83': 0.3129049481,
 'feat_84': 0.1182504766,
 'feat_85': 0,
 'feat_86': 0.7168453432,
 'feat_87': 0.00921944,
 'feat_88': 0.6228789594,
 'feat_89': 0.6014532901,
 'feat_90': 0.827890214,
 'feat_91': 0.2902018928,
 'feat_92': 0.0266006308,
 'feat_93': 0.5640501123,
 'feat_94': 1,
 'feat_95': 0.0164687409}
In [72]:
len(taiwan_data["observations"][0])
Out[72]:
97
In [73]:
# Calculate the number of features associated with each company and assign the result to n_features
n_features = len(taiwan_data["observations"][0])
print(n_features)

97

I can now create a wrangle function that takes as input the path of a compressed JSON file and returns the file's contents as a DataFrame. The index of the DataFrame contains the ID of the companies. When my function is complete, I'll use it to load the data into the DataFrame df.

In [74]:
# Iterate through companies
for item in taiwan_data["observations"]:
    if len(item) !=97:
        print("ALERT!!")
In [75]:
# Open compressed file and load contents
with gzip.open("data/taiwan-bankruptcy-data.json.gz", "r") as read_file:
    taiwan_data_gz = json.load(read_file)

print(type(taiwan_data_gz))

<class 'dict'>
In [76]:
# Explore taiwan_data_gz`
print(taiwan_data_gz.keys())
print(len(taiwan_data_gz["observations"]))
print(len(taiwan_data_gz["observations"][0]))

dict_keys(['schema', 'metadata', 'observations'])
6137
97
In [77]:
df = pd.DataFrame().from_dict(taiwan_data_gz["observations"]).set_index("id")
print(df.shape)
df.head()

(6137, 96)
Out[77]:
bankrupt feat_1 feat_2 feat_3 feat_4 feat_5 feat_6 feat_7 feat_8 feat_9 ... feat_86 feat_87 feat_88 feat_89 feat_90 feat_91 feat_92 feat_93 feat_94 feat_95
id
1 True 0.370594 0.424389 0.405750 0.601457 0.601457 0.998969 0.796887 0.808809 0.302646 ... 0.716845 0.009219 0.622879 0.601453 0.827890 0.290202 0.026601 0.564050 1 0.016469
2 True 0.464291 0.538214 0.516730 0.610235 0.610235 0.998946 0.797380 0.809301 0.303556 ... 0.795297 0.008323 0.623652 0.610237 0.839969 0.283846 0.264577 0.570175 1 0.020794
3 True 0.426071 0.499019 0.472295 0.601450 0.601364 0.998857 0.796403 0.808388 0.302035 ... 0.774670 0.040003 0.623841 0.601449 0.836774 0.290189 0.026555 0.563706 1 0.016474
4 True 0.399844 0.451265 0.457733 0.583541 0.583541 0.998700 0.796967 0.808966 0.303350 ... 0.739555 0.003252 0.622929 0.583538 0.834697 0.281721 0.026697 0.564663 1 0.023982
5 True 0.465022 0.538432 0.522298 0.598783 0.598783 0.998973 0.797366 0.809304 0.303475 ... 0.795016 0.003878 0.623521 0.598782 0.839973 0.278514 0.024752 0.575617 1 0.035490

5 rows × 96 columns

In [78]:
# Create wrangle function
def wrangle(filename):
    # Open compressed file, load into dict
    with gzip.open(filename, "r") as f:
        data = json.load(f)
        
    # Turn dict into DataFrame
    df = pd.DataFrame().from_dict(data["observations"]).set_index("id")
    
    return df
In [79]:
df = wrangle("data/taiwan-bankruptcy-data.json.gz")
print("df shape:", df.shape)
df.head()

df shape: (6137, 96)
Out[79]:
bankrupt feat_1 feat_2 feat_3 feat_4 feat_5 feat_6 feat_7 feat_8 feat_9 ... feat_86 feat_87 feat_88 feat_89 feat_90 feat_91 feat_92 feat_93 feat_94 feat_95
id
1 True 0.370594 0.424389 0.405750 0.601457 0.601457 0.998969 0.796887 0.808809 0.302646 ... 0.716845 0.009219 0.622879 0.601453 0.827890 0.290202 0.026601 0.564050 1 0.016469
2 True 0.464291 0.538214 0.516730 0.610235 0.610235 0.998946 0.797380 0.809301 0.303556 ... 0.795297 0.008323 0.623652 0.610237 0.839969 0.283846 0.264577 0.570175 1 0.020794
3 True 0.426071 0.499019 0.472295 0.601450 0.601364 0.998857 0.796403 0.808388 0.302035 ... 0.774670 0.040003 0.623841 0.601449 0.836774 0.290189 0.026555 0.563706 1 0.016474
4 True 0.399844 0.451265 0.457733 0.583541 0.583541 0.998700 0.796967 0.808966 0.303350 ... 0.739555 0.003252 0.622929 0.583538 0.834697 0.281721 0.026697 0.564663 1 0.023982
5 True 0.465022 0.538432 0.522298 0.598783 0.598783 0.998973 0.797366 0.809304 0.303475 ... 0.795016 0.003878 0.623521 0.598782 0.839973 0.278514 0.024752 0.575617 1 0.035490

5 rows × 96 columns

Explore¶

I want to see if there any missing data in the dataset. I can then create a Series where the index contains the name of the columns in df and the values are the number of NaNs in each column. Then I will assign the result to nans_by_col. Neither the Series itself nor its index require a name.

In [80]:
df.info()

<class 'pandas.core.frame.DataFrame'>
Int64Index: 6137 entries, 1 to 6819
Data columns (total 96 columns):
 #   Column    Non-Null Count  Dtype  
---  ------    --------------  -----  
 0   bankrupt  6137 non-null   bool   
 1   feat_1    6137 non-null   float64
 2   feat_2    6137 non-null   float64
 3   feat_3    6137 non-null   float64
 4   feat_4    6137 non-null   float64
 5   feat_5    6137 non-null   float64
 6   feat_6    6137 non-null   float64
 7   feat_7    6137 non-null   float64
 8   feat_8    6137 non-null   float64
 9   feat_9    6137 non-null   float64
 10  feat_10   6137 non-null   float64
 11  feat_11   6137 non-null   float64
 12  feat_12   6137 non-null   float64
 13  feat_13   6137 non-null   float64
 14  feat_14   6137 non-null   float64
 15  feat_15   6137 non-null   float64
 16  feat_16   6137 non-null   float64
 17  feat_17   6137 non-null   float64
 18  feat_18   6137 non-null   float64
 19  feat_19   6137 non-null   float64
 20  feat_20   6137 non-null   float64
 21  feat_21   6137 non-null   float64
 22  feat_22   6137 non-null   float64
 23  feat_23   6137 non-null   float64
 24  feat_24   6137 non-null   float64
 25  feat_25   6137 non-null   float64
 26  feat_26   6137 non-null   float64
 27  feat_27   6137 non-null   float64
 28  feat_28   6137 non-null   float64
 29  feat_29   6137 non-null   float64
 30  feat_30   6137 non-null   float64
 31  feat_31   6137 non-null   float64
 32  feat_32   6137 non-null   float64
 33  feat_33   6137 non-null   float64
 34  feat_34   6137 non-null   float64
 35  feat_35   6137 non-null   float64
 36  feat_36   6137 non-null   float64
 37  feat_37   6137 non-null   float64
 38  feat_38   6137 non-null   float64
 39  feat_39   6137 non-null   float64
 40  feat_40   6137 non-null   float64
 41  feat_41   6137 non-null   float64
 42  feat_42   6137 non-null   float64
 43  feat_43   6137 non-null   float64
 44  feat_44   6137 non-null   float64
 45  feat_45   6137 non-null   float64
 46  feat_46   6137 non-null   float64
 47  feat_47   6137 non-null   float64
 48  feat_48   6137 non-null   float64
 49  feat_49   6137 non-null   float64
 50  feat_50   6137 non-null   float64
 51  feat_51   6137 non-null   float64
 52  feat_52   6137 non-null   float64
 53  feat_53   6137 non-null   float64
 54  feat_54   6137 non-null   float64
 55  feat_55   6137 non-null   float64
 56  feat_56   6137 non-null   float64
 57  feat_57   6137 non-null   float64
 58  feat_58   6137 non-null   float64
 59  feat_59   6137 non-null   float64
 60  feat_60   6137 non-null   float64
 61  feat_61   6137 non-null   float64
 62  feat_62   6137 non-null   float64
 63  feat_63   6137 non-null   float64
 64  feat_64   6137 non-null   float64
 65  feat_65   6137 non-null   float64
 66  feat_66   6137 non-null   float64
 67  feat_67   6137 non-null   float64
 68  feat_68   6137 non-null   float64
 69  feat_69   6137 non-null   float64
 70  feat_70   6137 non-null   float64
 71  feat_71   6137 non-null   float64
 72  feat_72   6137 non-null   float64
 73  feat_73   6137 non-null   float64
 74  feat_74   6137 non-null   float64
 75  feat_75   6137 non-null   float64
 76  feat_76   6137 non-null   float64
 77  feat_77   6137 non-null   float64
 78  feat_78   6137 non-null   float64
 79  feat_79   6137 non-null   float64
 80  feat_80   6137 non-null   float64
 81  feat_81   6137 non-null   float64
 82  feat_82   6137 non-null   float64
 83  feat_83   6137 non-null   float64
 84  feat_84   6137 non-null   float64
 85  feat_85   6137 non-null   int64  
 86  feat_86   6137 non-null   float64
 87  feat_87   6137 non-null   float64
 88  feat_88   6137 non-null   float64
 89  feat_89   6137 non-null   float64
 90  feat_90   6137 non-null   float64
 91  feat_91   6137 non-null   float64
 92  feat_92   6137 non-null   float64
 93  feat_93   6137 non-null   float64
 94  feat_94   6137 non-null   int64  
 95  feat_95   6137 non-null   float64
dtypes: bool(1), float64(93), int64(2)
memory usage: 4.5 MB
In [81]:
df.isnull().sum().sum()
Out[81]:
0
In [82]:
df.nunique()
Out[82]:
bankrupt       2
feat_1      3159
feat_2      2985
feat_3      2984
feat_4      3580
            ... 
feat_91     6136
feat_92     5599
feat_93     5608
feat_94        1
feat_95     6137
Length: 96, dtype: int64
In [83]:
df.isna().sum()
Out[83]:
bankrupt    0
feat_1      0
feat_2      0
feat_3      0
feat_4      0
           ..
feat_91     0
feat_92     0
feat_93     0
feat_94     0
feat_95     0
Length: 96, dtype: int64
In [84]:
nans_by_col = pd.Series(df.isna().sum())
print("nans_by_col shape:", nans_by_col.shape)
nans_by_col.head()

nans_by_col shape: (96,)
Out[84]:
bankrupt    0
feat_1      0
feat_2      0
feat_3      0
feat_4      0
dtype: int64

To see if the data imbalanced, I can create a bar chart that shows the normalized value counts for the column df["bankrupt"].

In [85]:
df["bankrupt"].value_counts()
Out[85]:
False    5947
True      190
Name: bankrupt, dtype: int64
In [86]:
df["bankrupt"].value_counts(normalize=True)
Out[86]:
False    0.96904
True     0.03096
Name: bankrupt, dtype: float64
In [87]:
# Plot class balance
df["bankrupt"].value_counts(normalize=True).plot(
    kind="bar",
    xlabel="Bankrupt",  # Label x-axis "Bankrupt"
    ylabel="Frequency", # Label y-axis "Frequency"
    title="Class Balance" # Title "Chart Balance"
);

Split¶

In [88]:
# Create my feature matrix X and target vector y
target = "bankrupt" # My target is "bankrupt"
X = df.drop(columns="bankrupt") # Drop column with multicollinearity in the feature matrix
y = df[target]

print("X shape:", X.shape)
print("y shape:", y.shape)

X shape: (6137, 95)
y shape: (6137,)
In [89]:
# Divide my dataset into training and test sets using a randomized split
X_train, X_test, y_train, y_test =  train_test_split(
    X, y, test_size=0.2, random_state=42   # My test set should be 20% of my data & set random_state to 42
)
print("X_train shape:", X_train.shape)
print("y_train shape:", y_train.shape)
print("X_test shape:", X_test.shape)
print("y_test shape:", y_test.shape)

X_train shape: (4909, 95)
y_train shape: (4909,)
X_test shape: (1228, 95)
y_test shape: (1228,)

Resample¶

In [90]:
# Create a new feature matrix X_train_over and target vector y_train_over by performing random over-sampling on the training data
over_sampler = RandomOverSampler(random_state=42) # Set the random_state to 42
X_train_over, y_train_over = over_sampler.fit_resample(X_train, y_train)
print("X_train_over shape:", X_train_over.shape)
X_train_over.head()

X_train_over shape: (9512, 95)
Out[90]:
feat_1 feat_2 feat_3 feat_4 feat_5 feat_6 feat_7 feat_8 feat_9 feat_10 ... feat_86 feat_87 feat_88 feat_89 feat_90 feat_91 feat_92 feat_93 feat_94 feat_95
0 0.535855 0.599160 0.594411 0.627099 0.627099 0.999220 0.797686 0.809591 0.303518 0.781865 ... 0.834091 0.022025 0.624364 0.627101 0.841977 0.275384 0.026791 0.565158 1 0.147943
1 0.554136 0.612734 0.595000 0.607388 0.607388 0.999120 0.797614 0.809483 0.303600 0.781754 ... 0.840293 0.002407 0.624548 0.607385 0.842645 0.276532 0.026791 0.565158 1 0.062544
2 0.549554 0.603467 0.599122 0.620166 0.620166 0.999119 0.797569 0.809470 0.303524 0.781740 ... 0.840403 0.000840 0.624010 0.620163 0.842873 0.277249 0.026800 0.565200 1 0.047929
3 0.543801 0.603249 0.606992 0.622515 0.622515 0.999259 0.797728 0.809649 0.303510 0.781930 ... 0.831514 0.006176 0.626775 0.622513 0.842989 0.280013 0.026839 0.565375 1 0.028386
4 0.498659 0.562364 0.546978 0.603670 0.603670 0.998904 0.797584 0.809459 0.304000 0.781713 ... 0.811988 0.004256 0.623674 0.603669 0.841105 0.277628 0.026897 0.565618 1 0.043080

5 rows × 95 columns

Build Model¶

Baseline¶

In [91]:
# Calculate the baseline accuracy score for my model
acc_baseline = y_train.value_counts(normalize=True).max()
print("Baseline Accuracy:", round(acc_baseline, 4))

Baseline Accuracy: 0.9688

Iterate¶

Gradient Boosting Trees

In [92]:
# Create a classifier clf that can be trained on (X_train_over, y_train_over)
clf = RandomForestClassifier(random_state=42)
print(clf)

RandomForestClassifier(random_state=42)
In [93]:
# Perform cross-validation with my classifier using the over-sampled training data, and assign my results to cv_scores
cv_scores = cross_val_score(clf, X_train_over, y_train_over, cv=5, n_jobs=-1)  # Set the cv argument to 5
print(cv_scores)

[0.99316868 0.99474514 0.99369085 0.99369085 0.9957939 ]
In [94]:
# Create a dictionary params with the range of hyperparameters that I want to evaluate for my classifier
params = {
    "n_estimators": range(25, 100, 25),
    "max_depth": range(10, 50, 10)


}
params
Out[94]:
{'n_estimators': range(25, 100, 25), 'max_depth': range(10, 50, 10)}
In [95]:
# Create a GridSearchCV named model that includes my classifier and hyperparameter grid
model = GridSearchCV(clf, param_grid=params, cv=5, n_jobs=-1, verbose=1) # Set cv to 5, n_jobs to -1, and verbose to 1
model
Out[95]:
GridSearchCV(cv=5, estimator=RandomForestClassifier(random_state=42), n_jobs=-1,
             param_grid={'max_depth': range(10, 50, 10),
                         'n_estimators': range(25, 100, 25)},
             verbose=1)
In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
GridSearchCV(cv=5, estimator=RandomForestClassifier(random_state=42), n_jobs=-1,
             param_grid={'max_depth': range(10, 50, 10),
                         'n_estimators': range(25, 100, 25)},
             verbose=1)
RandomForestClassifier(random_state=42)
RandomForestClassifier(random_state=42)
In [96]:
# Fit my model to the over-sampled training data
model.fit(X_train_over, y_train_over)

Fitting 5 folds for each of 12 candidates, totalling 60 fits
Out[96]:
GridSearchCV(cv=5, estimator=RandomForestClassifier(random_state=42), n_jobs=-1,
             param_grid={'max_depth': range(10, 50, 10),
                         'n_estimators': range(25, 100, 25)},
             verbose=1)
In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
GridSearchCV(cv=5, estimator=RandomForestClassifier(random_state=42), n_jobs=-1,
             param_grid={'max_depth': range(10, 50, 10),
                         'n_estimators': range(25, 100, 25)},
             verbose=1)
RandomForestClassifier(random_state=42)
RandomForestClassifier(random_state=42)
In [97]:
# Extract the cross-validation results from my model, and load them into a DataFrame named cv_results
cv_results = pd.DataFrame(model.cv_results_)
cv_results.head(5)
Out[97]:
mean_fit_time std_fit_time mean_score_time std_score_time param_max_depth param_n_estimators params split0_test_score split1_test_score split2_test_score split3_test_score split4_test_score mean_test_score std_test_score rank_test_score
0 0.466611 0.030509 0.006319 0.002242 10 25 {'max_depth': 10, 'n_estimators': 25} 0.981608 0.980032 0.979495 0.981073 0.980547 0.980551 0.000745 11
1 0.906622 0.022882 0.009819 0.002386 10 50 {'max_depth': 10, 'n_estimators': 50} 0.983184 0.981608 0.977918 0.980021 0.981073 0.980761 0.001750 10
2 1.409413 0.038556 0.013992 0.002892 10 75 {'max_depth': 10, 'n_estimators': 75} 0.983184 0.981083 0.977392 0.979495 0.981073 0.980445 0.001925 12
3 0.566776 0.029568 0.008576 0.002357 20 25 {'max_depth': 20, 'n_estimators': 25} 0.990541 0.986863 0.985804 0.990011 0.989485 0.988541 0.001863 9
4 1.125918 0.056730 0.014820 0.003020 20 50 {'max_depth': 20, 'n_estimators': 50} 0.990541 0.990016 0.987907 0.990011 0.990011 0.989697 0.000918 7
In [98]:
# Extract the best hyperparameters from my model and assign them to best_params
best_params = model.best_params_
print(best_params)

{'max_depth': 40, 'n_estimators': 75}

Evaluate¶

In [99]:
# Test the quality of my model by calculating accuracy scores for the training and test data
acc_train = model.score(X_train, y_train)
acc_test = model.score(X_test, y_test)

print("Model Training Accuracy:", round(acc_train, 4))
print("Model Test Accuracy:", round(acc_test, 4))

Model Training Accuracy: 1.0
Model Test Accuracy: 0.9764
In [100]:
# Plot a confusion matrix that shows how my model performed on my test set
ConfusionMatrixDisplay.from_estimator(model, X_test, y_test);
In [101]:
# Generate a classification report for my model's performance on the test data and assign it to class_report
class_report = classification_report(y_test, model.predict(X_test))
print(class_report)

              precision    recall  f1-score   support

       False       0.98      0.99      0.99      1191
        True       0.68      0.41      0.51        37

    accuracy                           0.98      1228
   macro avg       0.83      0.70      0.75      1228
weighted avg       0.97      0.98      0.97      1228

Communicate¶

I will create a horizontal bar chart with the 10 most important features for my model.

In [102]:
# Get feature names from training data
features = X_train_over.columns

# Extract importances from model
importances = model.best_estimator_.feature_importances_

# Create a series with feature names and importances
feat_imp = pd.Series(importances, index=features).sort_values()

# Plot 10 most important features
feat_imp.tail(10).plot(kind="barh")
plt.xlabel("Gini Importance")
plt.ylabel("Feature")
plt.title("Feature Importance");
In [103]:
# Save my best-performing model to a a file named "model-5-5.pkl"
with open("model-5-5.pkl", "wb") as f:
    pickle.dump(model, f)

Now, I can open the file my_predictor_taiwan.py. Add my wrangle function, and then create a make_predictions function that takes two arguments: data_filepath and model_filepath.

In [1]:
%%bash

cat my_predictor_taiwan.py

# Import libraries
import gzip
import json
import pickle
import pandas as pd

def wrangle(filename):
    # Open compressed file, load into dict
    with gzip.open(filename, "r") as f:
        data = json.load(f)
        
    # Turn dict into DataFrame
    df = pd.DataFrame().from_dict(data["observations"]).set_index("id")
    
    return df


def make_predictions(data_filepath, model_filepath):
    # Wrangle JSON file
    X_test = wrangle(data_filepath)
    # Load model
    with open(model_filepath, "rb") as f:
        model = pickle.load(f)
    # Generate predictions
    y_test_pred = model.predict(X_test)
    # Put predictions into Series w/ name "bankrupt", and same index as X_test
    y_test_pred = pd.Series(y_test_pred, index=X_test.index, name="bankrupt")
    return y_test_pred
In [111]:
# Import my module
from my_predictor_taiwan import make_predictions

# Generate predictions
y_test_pred = make_predictions(
    data_filepath="data/taiwan-bankruptcy-data-test-features.json.gz",
    model_filepath="model-5-5.pkl",
)

print("predictions shape:", y_test_pred.shape)
y_test_pred.head()

predictions shape: (682,)
Out[111]:
id
18    False
20    False
24    False
32    False
38    False
Name: bankrupt, dtype: bool