The LendingClub is a peer-to-peer leading company that directly connects borrowers and potential lenders/investors. In this notebook, you will build a classification model to predict whether or not a loan provided by LendingClub is likely to default).
In this notebook you will use data from the LendingClub to predict whether a loan will be paid off in full or the loan will be charged off and possibly go into default. In this assignment you will:
Let's get started!
In [42]:
import json
import numpy as np
import pandas as pd
import sklearn, sklearn.tree
import matplotlib as mpl
import matplotlib.pyplot as plt
import seaborn as sns
sns.set_style('darkgrid')
%matplotlib inline
We will be using a dataset from the LendingClub. A parsed and cleaned form of the dataset is availiable here. Make sure you download the dataset before running the following command.
In [43]:
loans = pd.read_csv("lending-club-data_assign_1.csv")
In [44]:
loans.head()
Out[44]:
Now, let's print out the column names to see what features we have in this dataset.
In [45]:
loans.columns.values
Out[45]:
Here, we see that we have some feature columns that have to do with grade of the loan, annual income, home ownership status, etc. Let's take a look at the distribution of loan grades in the dataset.
In [46]:
plt.figure(figsize=(10,6))
loans['grade'].value_counts().plot(kind='bar')
plt.tick_params(axis='x', labelsize=18)
plt.xticks(rotation='horizontal')
plt.tick_params(axis='y', labelsize=18)
plt.title("Histogram of Loan Grades", fontsize=18)
plt.xlabel("Loan Grades", fontsize=18)
plt.ylabel("Count", fontsize=18)
Out[46]:
We can see that over half of the loan grades are assigned values B
or C
. Each loan is assigned one of these grades, along with a more finely discretized feature called sub_grade
(feel free to explore that feature column as well!). These values depend on the loan application and credit report, and determine the interest rate of the loan. More information can be found here.
Now, let's look at a different feature.
In [47]:
plt.figure(figsize=(10,6))
loans['home_ownership'].value_counts().plot(kind='bar')
plt.tick_params(axis='x', labelsize=18)
plt.xticks(rotation='horizontal')
plt.tick_params(axis='y', labelsize=18)
plt.title("Histogram of Home Ownership", fontsize=18)
plt.xlabel("Home Ownership Type", fontsize=18)
plt.ylabel("Count", fontsize=18)
Out[47]:
This feature describes whether the loanee is mortaging, renting, or owns a home. We can see that a small percentage of the loanees own a home.
The target column (label column) of the dataset that we are interested in is called bad_loans
. In this column 1 means a risky (bad) loan 0 means a safe loan.
In order to make this more intuitive and consistent with the lectures, we reassign the target to be:
We put this in a new column called safe_loans
.
In [48]:
# safe_loans = 1 => safe
# safe_loans = -1 => risky
loans['safe_loans'] = loans['bad_loans'].apply(lambda x : +1 if x==0 else -1)
loans = loans.drop('bad_loans', 1)
Now, let us explore the distribution of the column safe_loans
. This gives us a sense of how many safe and risky loans are present in the dataset.
In [49]:
plt.figure(figsize=(10,6))
loans['safe_loans'].value_counts().plot(kind='bar')
plt.tick_params(axis='x', labelsize=18)
plt.xticks(rotation='horizontal')
plt.tick_params(axis='y', labelsize=18)
plt.title("Histogram of whether a Loan is safe or risky", fontsize=18)
plt.xlabel("Safe Loan=1, Risky Loan=-1", fontsize=18)
plt.ylabel("Count", fontsize=18)
Out[49]:
In [50]:
print "Percentage of safe loans: %.1f%%" %((loans['safe_loans'].value_counts().ix[1]/float(len(loans['safe_loans'])))*100.0)
print "Percentage of risky loans: %.1f%%" %((loans['safe_loans'].value_counts().ix[-1]/float(len(loans['safe_loans'])))*100.0)
You should have:
It looks like most of these loans are safe loans (thankfully). But this does make our problem of identifying risky loans challenging.
In this assignment, we will be using a subset of features (categorical and numeric). The features we will be using are described in the code comments below. If you are a finance geek, the LendingClub website has a lot more details about these features.
In [51]:
features = ['grade', # grade of the loan
'sub_grade', # sub-grade of the loan
'short_emp', # one year or less of employment
'emp_length_num', # number of years of employment
'home_ownership', # home_ownership status: own, mortgage or rent
'dti', # debt to income ratio
'purpose', # the purpose of the loan
'term', # the term of the loan
'last_delinq_none', # has borrower had a delinquincy
'last_major_derog_none', # has borrower had 90 day or worse rating
'revol_util', # percent of available credit being used
'total_rec_late_fee', # total late fees received to day
]
target = 'safe_loans' # prediction target (y) (+1 means safe, -1 is risky)
# Extract the feature columns and target column
loans = loans[features + [target]]
What remains now is a subset of features and the target that we will use for the rest of this notebook.
In [52]:
safe_loans_raw = loans[loans[target] == +1]
risky_loans_raw = loans[loans[target] == -1]
print "Number of safe loans : %s" % len(safe_loans_raw)
print "Number of risky loans : %s" % len(risky_loans_raw)
Now, write some code to compute below the percentage of safe and risky loans in the dataset and validate these numbers against what was calculated earlier in the assignment:
In [53]:
print "Percentage of safe loans : %.1f%%" %((float(len(safe_loans_raw))/len(loans[target]))*100.0)
print "Percentage of risky loans : %.1f%%" %((float(len(risky_loans_raw))/len(loans[target]))*100.0)
As can be seem there are much more sage loans than risky loans in the data set. The training data and validation data we will load will combat this class imbalance and will have roughly 50% safe loans and 50% risky loans.
In [54]:
loans_one_hot_enc = pd.get_dummies(loans)
Loading the JSON files with the indicies from the training data and the validation data into a a list.
In [55]:
with open('module-5-assignment-1-train-idx.json', 'r') as f:
train_idx_lst = json.load(f)
train_idx_lst = [int(entry) for entry in train_idx_lst]
In [56]:
with open('module-5-assignment-1-validation-idx.json', 'r') as f:
validation_idx_lst = json.load(f)
validation_idx_lst = [int(entry) for entry in validation_idx_lst]
Using the list of the training data indicies and the validation data indicies to get a DataFrame with the training data and a DataFrame with the validation data.
In [57]:
train_data = loans_one_hot_enc.ix[train_idx_lst]
validation_data = loans_one_hot_enc.ix[validation_idx_lst]
Now, let's use the built-in GraphLab Create decision tree learner to create a loan prediction model on the training data. (In the next assignment, you will implement your own decision tree learning algorithm.) Our feature columns and target column have already been decided above. Use validation_set=None
to get the same results as everyone else.
Using sklearn to learn a decision tree classification model. The first entry in .fit is all the data, excluding the target variable "safe_loans" and the second entry is the targer variable "safe_loans".
First, training a tree with max_depth=6
In [58]:
decision_tree_model = sklearn.tree.DecisionTreeClassifier(max_depth=6)
decision_tree_model.fit(train_data.ix[:, train_data.columns != "safe_loans"], train_data["safe_loans"])
Out[58]:
Now, training a tree with max_depth=2
In [59]:
small_model = sklearn.tree.DecisionTreeClassifier(max_depth=2)
small_model.fit(train_data.ix[:, train_data.columns != "safe_loans"], train_data["safe_loans"])
Out[59]:
In [60]:
validation_safe_loans = validation_data[validation_data[target] == 1]
validation_risky_loans = validation_data[validation_data[target] == -1]
sample_validation_data_risky = validation_risky_loans[0:2]
sample_validation_data_safe = validation_safe_loans[0:2]
sample_validation_data = sample_validation_data_safe.append(sample_validation_data_risky)
sample_validation_data
Out[60]:
Now, we will use our model to predict whether or not a loan is likely to default. For each row in the sample_validation_data, use the decision_tree_model to predict whether or not the loan is classified as a safe loan.
Hint: Be sure to use the .predict()
method.
In [61]:
samp_vald_data_pred = decision_tree_model.predict(sample_validation_data.ix[:, sample_validation_data.columns != "safe_loans"])
In [62]:
samp_vald_data_label = sample_validation_data["safe_loans"].values
Quiz Question: What percentage of the predictions on sample_validation_data
did decision_tree_model
get correct?
In [63]:
print "%.1f%%" %((np.sum(samp_vald_data_pred == samp_vald_data_label)/float(len(samp_vald_data_pred)))*100.0)
In [64]:
samp_vald_data_prob = decision_tree_model.predict_proba(sample_validation_data.ix[:, sample_validation_data.columns != "safe_loans"])[:,1]
Quiz Question: Which loan has the highest probability of being classified as a safe loan?
In [65]:
sample_validation_data.index[np.argmax(samp_vald_data_prob)]
Out[65]:
In [66]:
sample_validation_data
Out[66]:
41 corresponds to the 4th loan
Now, we will explore something pretty interesting. For each row in the sample_validation_data, what is the probability (according to small_model) of a loan being classified as safe?
Hint: Set output_type='probability'
to make probability predictions using small_model on sample_validation_data
:
In [67]:
small_model.predict_proba(sample_validation_data.ix[:, sample_validation_data.columns != "safe_loans"])[:,1]
Out[67]:
Quiz Question: Notice that the probability preditions are the exact same for the 2nd and 3rd loans. Why would this happen?
During tree traversal both examples fall into the same leaf node.
Now, let's consider the 2nd entry in the sample_validation_data
In [68]:
sample_validation_data
Out[68]:
The 2nd entry of sample_validation_data has index 79
In [69]:
sample_validation_data.ix[79]
Out[69]:
Quiz Question: Based on the small_model , what prediction would you make for this data point?
In [70]:
small_model.predict(sample_validation_data.ix[79, sample_validation_data.columns != "safe_loans"])[0]
Out[70]:
Recall that the accuracy is defined as follows: $$ \mbox{accuracy} = \frac{\mbox{# correctly classified examples}}{\mbox{# total examples}} $$
Let us start by evaluating the accuracy of the small_model
and decision_tree_model
on the training data
In [71]:
small_model_train_acc = small_model.score(train_data.ix[:, train_data.columns != "safe_loans"], train_data["safe_loans"])
decision_tree_model_train_acc = decision_tree_model.score(train_data.ix[:, train_data.columns != "safe_loans"], train_data["safe_loans"])
print small_model_train_acc
print decision_tree_model_train_acc
Checkpoint: You should see that the small_model performs worse than the decision_tree_model on the training data.
In [72]:
decision_tree_model_train_acc > small_model_train_acc
Out[72]:
Now, let us evaluate the accuracy of the small_model and decision_tree_model on the entire validation_data, not just the subsample considered above.
Quiz Question: What is the accuracy of decision_tree_model
on the validation set, rounded to the nearest .01?
In [73]:
decision_tree_model_valid_acc = decision_tree_model.score(validation_data.ix[:, validation_data.columns != "safe_loans"], validation_data["safe_loans"])
In [74]:
print "Accuracy of decision_tree_model on validation set: %.2f" %(decision_tree_model_valid_acc)
Here, we will train a large decision tree with max_depth=10
. This will allow the learned tree to become very deep, and result in a very complex model. Recall that in lecture, we prefer simpler models with similar predictive power. This will be an example of a more complicated model which has similar predictive power, i.e. something we don't want.
In [75]:
big_model = sklearn.tree.DecisionTreeClassifier(max_depth=10)
big_model.fit(train_data.ix[:, train_data.columns != "safe_loans"], train_data["safe_loans"])
Out[75]:
Now, let us evaluate the accuracy of the big_model on the training set and validation set.
In [76]:
big_model_train_acc = big_model.score(train_data.ix[:, train_data.columns != "safe_loans"], train_data["safe_loans"])
big_model_valid_acc = big_model.score(validation_data.ix[:, validation_data.columns != "safe_loans"], validation_data["safe_loans"])
Checkpoint: We should see that big_model has even better performance on the training set than decision_tree_model did on the training set.
In [77]:
big_model_train_acc > decision_tree_model_train_acc
Out[77]:
Quiz Question: How does the performance of big_model on the validation set compare to decision_tree_model on the validation set? Is this a sign of overfitting?
In [78]:
big_model_valid_acc > decision_tree_model_valid_acc
Out[78]:
The big_model has more features, performs better on the training dataset, but worse on the validation dataset. This is a sign of overfitting.
Every mistake the model makes costs money. In this section, we will try and quantify the cost of each mistake made by the model.
Assume the following:
Let's write code that can compute the cost of mistakes made by the model.
First, let us make predictions on validation_data
using the decision_tree_model
. Then, let's store the labels of validation_data
.
In [79]:
predic_valid_data = decision_tree_model.predict(validation_data.ix[:, validation_data.columns != "safe_loans"])
labels_valid_data = validation_data["safe_loans"].values
Now, let's initialize counters that will store the number of false positive and the number of false negatives to 0.
In [80]:
N_false_pos = 0
N_false_neg = 0
Now, let's loop over the data to determine the number of false positive and the number of false negatives. False positives are predictions where the model predicts +1 but the true label is -1. False negatives are predictions where the model predicts -1 but the true label is +1.
In [81]:
for i in range(len(labels_valid_data)):
# If we find a mistake
if predic_valid_data[i] != labels_valid_data[i]:
# If false positive, increment N_false_pos
if predic_valid_data[i]==1:
N_false_pos += 1
# Else, it's a false negative, increment N_false_neg
else:
N_false_neg += 1
Quiz Question: Let us assume that each mistake costs money:
What is the total cost of mistakes made by decision_tree_model
on validation_data
?
In [82]:
10000*N_false_neg + 20000*N_false_pos
Out[82]:
In [ ]: