def find_optimum_credit_score(
foreclosure_model, foreclosure_scaler, borrower_credit_score,
debt_to_income_ratio, lender, interest_rate, loan_amount, state,
zip_code, loan_term, loan_to_value, combined_loan_to_value,
borrower_count, first_time_homebuyer, loan_purpose, property_type,
unit_count, occupancy_status, product_type, co_borrower_credit_score,
income, low, high):
#print("current low credit= "+ str(low)+", current high credit= "+str(high))
if (low > high):
#print('checkpoint 1')
#print('-1, -1')
return -1, -1
elif (high - low <= 10):
#print('checkpoint 2')
foreclosure_probability = predict_single_data(
foreclosure_model, foreclosure_scaler, high, debt_to_income_ratio,
lender, interest_rate, loan_amount, state, zip_code, loan_term,
loan_to_value, combined_loan_to_value, borrower_count,
first_time_homebuyer, loan_purpose, property_type, unit_count,
occupancy_status, product_type, co_borrower_credit_score)
#print('foreclosure probability for credit score '+str(low)+' is '+str(foreclosure_probability))
if (get_pred_Status(foreclosure_probability) == 1):
#print('-1, -1')
return -1, -1
else:
#print("low credit = "+ str(low)+", high credit = "+str(high))
return low, high
#print('checkpoint 3')
mid = low + (high - low) / 2
foreclosure_probability = predict_single_data(
foreclosure_model, foreclosure_scaler, mid, debt_to_income_ratio,
lender, interest_rate, loan_amount, state, zip_code, loan_term,
loan_to_value, combined_loan_to_value, borrower_count,
first_time_homebuyer, loan_purpose, property_type, unit_count,
occupancy_status, product_type, co_borrower_credit_score)
#print('foreclosure probability for credit '+str(mid)+' is '+str(foreclosure_probability))
if (get_pred_Status(foreclosure_probability) == 1):
return find_optimum_credit_score(
foreclosure_model, foreclosure_scaler, borrower_credit_score,
debt_to_income_ratio, lender, interest_rate, loan_amount, state,
zip_code, loan_term, loan_to_value, combined_loan_to_value,
borrower_count, first_time_homebuyer, loan_purpose, property_type,
unit_count, occupancy_status, product_type,
co_borrower_credit_score, income, mid, high)
else:
return find_optimum_credit_score(
foreclosure_model, foreclosure_scaler, borrower_credit_score,
debt_to_income_ratio, lender, interest_rate, loan_amount, state,
zip_code, loan_term, loan_to_value, combined_loan_to_value,
borrower_count, first_time_homebuyer, loan_purpose, property_type,
unit_count, occupancy_status, product_type,
co_borrower_credit_score, income, low, mid)
def find_optimum_bank(foreclosure_model, foreclosure_scaler,
borrower_credit_score, debt_to_income_ratio, lender,
interest_rate, loan_amount, state, zip_code, loan_term,
loan_to_value, combined_loan_to_value, borrower_count,
first_time_homebuyer, loan_purpose, property_type,
unit_count, occupancy_status, product_type,
co_borrower_credit_score):
temp = list(ALL_SELECTION_OTIONS["lender"].items())
num_banks = len(temp)
for i in range(num_banks):
foreclosure_probability = predict_single_data(
foreclosure_model, foreclosure_scaler, borrower_credit_score,
debt_to_income_ratio, i, interest_rate, loan_amount, state,
zip_code, loan_term, loan_to_value, combined_loan_to_value,
borrower_count, first_time_homebuyer, loan_purpose, property_type,
unit_count, occupancy_status, product_type,
co_borrower_credit_score)
def find_optimum_salary(foreclosure_model, foreclosure_scaler,
borrower_credit_score, debt_to_income_ratio, lender,
interest_rate, loan_amount, state, zip_code, loan_term,
loan_to_value, combined_loan_to_value, borrower_count,
first_time_homebuyer, loan_purpose, property_type,
unit_count, occupancy_status, product_type,
co_borrower_credit_score, low, high):
if (low > high):
return -1, -1
elif (high - low <= 2000):
new_debt_to_income_ratio = compute_dti(high, loan_amount,
interest_rate, loan_term)
foreclosure_probability = predict_single_data(
foreclosure_model, foreclosure_scaler, borrower_credit_score,
new_debt_to_income_ratio, lender, interest_rate, loan_amount,
state, zip_code, loan_term, loan_to_value, loan_to_value,
borrower_count, first_time_homebuyer, loan_purpose, property_type,
unit_count, occupancy_status, product_type,
co_borrower_credit_score)
if (get_pred_Status(foreclosure_probability) == 1):
print('will return -1: foreclosure probabiloity for salary ' +
str(high) + 'is ' + str(foreclosure_probability))
return -1, -1
else:
return low, high
mid = low + (high - low) / 2
new_debt_to_income_ratio = compute_dti(mid, loan_amount, interest_rate,
loan_term)
foreclosure_probability = predict_single_data(
foreclosure_model, foreclosure_scaler, borrower_credit_score,
new_debt_to_income_ratio, lender, interest_rate, loan_amount, state,
zip_code, loan_term, loan_to_value, loan_to_value, borrower_count,
first_time_homebuyer, loan_purpose, property_type, unit_count,
occupancy_status, product_type, co_borrower_credit_score)
print('foreclosure probabiloity for salary ' + str(mid) + 'is ' +
str(foreclosure_probability))
if (get_pred_Status(foreclosure_probability) == 1):
return find_optimum_salary(
foreclosure_model, foreclosure_scaler, borrower_credit_score,
new_debt_to_income_ratio, lender, interest_rate, loan_amount,
state, zip_code, loan_term, loan_to_value, combined_loan_to_value,
borrower_count, first_time_homebuyer, loan_purpose, property_type,
unit_count, occupancy_status, product_type,
co_borrower_credit_score, mid, high)
else:
return find_optimum_salary(
foreclosure_model, foreclosure_scaler, borrower_credit_score,
new_debt_to_income_ratio, lender, interest_rate, loan_amount,
state, zip_code, loan_term, loan_to_value, combined_loan_to_value,
borrower_count, first_time_homebuyer, loan_purpose, property_type,
unit_count, occupancy_status, product_type,
co_borrower_credit_score, low, mid)
def create_array_for_chart(param, low, high, foreclosure_model,
foreclosure_scaler, borrower_credit_score,
debt_to_income_ratio, lender, interest_rate,
loan_amount, state, zip_code, loan_term,
loan_to_value, combined_loan_to_value,
borrower_count, first_time_homebuyer, loan_purpose,
property_type, unit_count, occupancy_status,
product_type, co_borrower_credit_score, income):
value_array = []
prob_list = []
if (param == 'interest_rate'):
while (low <= high):
new_debt_to_income_ratio = compute_dti(income, loan_amount, low,
loan_term)
foreclosure_probability = predict_single_data(
foreclosure_model, foreclosure_scaler, borrower_credit_score,
new_debt_to_income_ratio, lender, low, loan_amount, state,
zip_code, loan_term, loan_to_value, combined_loan_to_value,
borrower_count, first_time_homebuyer, loan_purpose,
property_type, unit_count, occupancy_status, product_type,
co_borrower_credit_score)
value_array.append(low)
prob_list.append(foreclosure_probability)
low = low + 0.05
elif (param == 'property_value'):
while (low <= high):
down_payment = 100.0 - loan_to_value
new_loan_amount = (low * (100.0 - down_payment)) / 100.0
new_debt_to_income_ratio = compute_dti(income, new_loan_amount,
interest_rate, loan_term)
foreclosure_probability = predict_single_data(
foreclosure_model, foreclosure_scaler, borrower_credit_score,
new_debt_to_income_ratio, lender, interest_rate,
new_loan_amount, state, zip_code, loan_term, loan_to_value,
combined_loan_to_value, borrower_count, first_time_homebuyer,
loan_purpose, property_type, unit_count, occupancy_status,
product_type, co_borrower_credit_score)
value_array.append(low)
prob_list.append(foreclosure_probability)
low = low + 5000
elif (param == 'borrower_credit_score'):
while (low <= high):
foreclosure_probability = predict_single_data(
foreclosure_model, foreclosure_scaler, high,
debt_to_income_ratio, lender, interest_rate, loan_amount,
state, zip_code, loan_term, loan_to_value,
combined_loan_to_value, borrower_count, first_time_homebuyer,
loan_purpose, property_type, unit_count, occupancy_status,
product_type, co_borrower_credit_score)
value_array.append(high)
prob_list.append(foreclosure_probability)
high = high - 5
elif (param == 'down_payment'):
while (low <= high):
new_loan_to_value = 100.0 - high
new_loan_amount = 1.0 * loan_amount * new_loan_to_value / loan_to_value
new_debt_to_income_ratio = compute_dti(income, new_loan_amount,
interest_rate, loan_term)
foreclosure_probability = predict_single_data(
foreclosure_model, foreclosure_scaler, borrower_credit_score,
new_debt_to_income_ratio, lender, interest_rate,
new_loan_amount, state, zip_code, loan_term, new_loan_to_value,
new_loan_to_value, borrower_count, first_time_homebuyer,
loan_purpose, property_type, unit_count, occupancy_status,
product_type, co_borrower_credit_score)
value_array.append(high)
prob_list.append(foreclosure_probability)
high = high - 1
elif (param == 'salary'):
while (low <= high):
new_debt_to_income_ratio = compute_dti(high, loan_amount,
interest_rate, loan_term)
foreclosure_probability = predict_single_data(
foreclosure_model, foreclosure_scaler, borrower_credit_score,
new_debt_to_income_ratio, lender, interest_rate, loan_amount,
state, zip_code, loan_term, loan_to_value, loan_to_value,
borrower_count, first_time_homebuyer, loan_purpose,
property_type, unit_count, occupancy_status, product_type,
co_borrower_credit_score)
value_array.append(high)
prob_list.append(foreclosure_probability)
high = high - 1000
print(value_array)
print(prob_list)
return value_array, prob_list
def predict():
#int_features = [x for x in request.form.values()]
#final = np.array(int_features)
#data_unseen = pd.DataFrame([final], columns = cols)
#prediction = predict_model(model, data=data_unseen, round = 0)
#prediction = int(prediction.Label[0])
#prediction =10
if 'annual_income' not in request.form:
return render_template('predict_page.html',
pred='',
options=ALL_SELECTION_OTIONS,
selections=get_default_selections(),
recommendations=[])
lender = get_categorical_value('lender', 'seller')
interest_rate = float(request.form['interest_rate'])
property_value = int(request.form['val'])
down_payment = int(request.form['down_payment'])
loan_amount = (property_value * (100 - down_payment)) / 100
loan_term = int(
request.form['loan_term']) * 12 if 'loan_term' in request.form else 360
loan_to_value = 100 - down_payment
combined_loan_to_value = loan_to_value
borrower_count = 2 if len(
request.form['co_borrower_credit_score']) > 0 and int(
request.form['co_borrower_credit_score']) > 0 else 1
income = int(request.form['annual_income'])
debt_to_income_ratio = compute_dti(income, loan_amount, interest_rate,
loan_term)
borrower_credit_score = int(request.form['borrower_credit_score'])
first_time_homebuyer = get_categorical_value('first_time_homebuyer',
'first_time_homebuyer')
loan_purpose = get_categorical_value('loan_purpose', 'loan_purpose')
property_type = get_categorical_value('property_type', 'property_type')
unit_count = int(
request.form['unit_count']) if 'unit_count' in request.form else 1
occupancy_status = get_categorical_value('occupancy_status',
'occupancy_status')
state = get_categorical_value('property_state', 'property_state')
zip_code = (int(request.form["zip"]) / 100)
product_type = 0
co_borrower_credit_score = int(
request.form['co_borrower_credit_score']) if len(
request.form['co_borrower_credit_score']) > 0 else 0
foreclosure_probability = predict_single_data(
foreclosure_model, foreclosure_scaler, borrower_credit_score,
debt_to_income_ratio, lender, interest_rate, loan_amount, state,
zip_code, loan_term, loan_to_value, combined_loan_to_value,
borrower_count, first_time_homebuyer, loan_purpose, property_type,
unit_count, occupancy_status, product_type, co_borrower_credit_score)
delinquency_probability = predict_single_data(
delinquency_model, delinquency_scaler, borrower_credit_score,
debt_to_income_ratio, lender, interest_rate, loan_amount, state,
zip_code, loan_term, loan_to_value, combined_loan_to_value,
borrower_count, first_time_homebuyer, loan_purpose, property_type,
unit_count, occupancy_status, product_type, co_borrower_credit_score)
#print('foreclosure_probability ({}) = {}', type(foreclosure_probability), foreclosure_probability)
#print('delinquency_probability ({}) = {}', type(delinquency_probability), delinquency_probability)
pred = form_prediction_result(foreclosure_probability,
delinquency_probability)
recommendations = []
if (get_pred_Status(foreclosure_probability) == 1):
result_property_low, result_property_high = find_optimum_price(
foreclosure_model, foreclosure_scaler, borrower_credit_score,
debt_to_income_ratio, lender, interest_rate, loan_amount / 2,
state, zip_code, loan_term, loan_to_value, combined_loan_to_value,
borrower_count, first_time_homebuyer, loan_purpose, property_type,
unit_count, occupancy_status, product_type,
co_borrower_credit_score, income, 40000, property_value)
print("prining optimum values for property_value")
print(result_property_low)
print(result_property_high)
if (result_property_low != -1):
value_array, prob_array = create_array_for_chart(
'property_value', result_property_low, property_value,
foreclosure_model, foreclosure_scaler, borrower_credit_score,
debt_to_income_ratio, lender, interest_rate, loan_amount,
state, zip_code, loan_term, loan_to_value,
combined_loan_to_value, borrower_count, first_time_homebuyer,
loan_purpose, property_type, unit_count, occupancy_status,
product_type, co_borrower_credit_score, income)
recommendations.append({
'param':
'property value',
'term':
'short',
'values':
value_array,
'probs':
prob_array,
'text':
'If you get a property of value ' + str(result_property_low) +
", your chance of foreclosure will go down to " +
str(int(100 * prob_array[0])) + '%.'
})
result_interest_rate_low, result_interest_rate_high = find_optimum_interest_rate(
foreclosure_model, foreclosure_scaler, borrower_credit_score,
debt_to_income_ratio, lender, interest_rate, loan_amount, state,
zip_code, loan_term, loan_to_value, combined_loan_to_value,
borrower_count, first_time_homebuyer, loan_purpose, property_type,
unit_count, occupancy_status, product_type,
co_borrower_credit_score, income, 2.3, interest_rate)
print("prining optimum values for interest_rate")
print(result_interest_rate_low)
print(result_interest_rate_high)
if (result_interest_rate_low != -1):
value_array, prob_array = create_array_for_chart(
'interest_rate', result_interest_rate_low, interest_rate,
foreclosure_model, foreclosure_scaler, borrower_credit_score,
debt_to_income_ratio, lender, interest_rate, loan_amount,
state, zip_code, loan_term, loan_to_value,
combined_loan_to_value, borrower_count, first_time_homebuyer,
loan_purpose, property_type, unit_count, occupancy_status,
product_type, co_borrower_credit_score, income)
recommendations.append({
'param':
'interest rate',
'term':
'short',
'values':
value_array,
'probs':
prob_array,
'text':
'If you get an interest rate of ' +
str("{:.2f}".format(result_interest_rate_low)) +
"%, your chance of foreclosure will go down to " +
str(int(100 * prob_array[0])) + '%.'
})
result_credit_score_low, result_credit_score_high = find_optimum_credit_score(
foreclosure_model, foreclosure_scaler, borrower_credit_score,
debt_to_income_ratio, lender, interest_rate, loan_amount, state,
zip_code, loan_term, loan_to_value, combined_loan_to_value,
borrower_count, first_time_homebuyer, loan_purpose, property_type,
unit_count, occupancy_status, product_type,
co_borrower_credit_score, income, borrower_credit_score, 850)
print("prining optimum values for borrower_credit_score")
print(result_credit_score_low)
print(result_credit_score_high)
if (result_credit_score_high != -1):
value_array, prob_array = create_array_for_chart(
'borrower_credit_score', borrower_credit_score,
result_credit_score_high, foreclosure_model,
foreclosure_scaler, borrower_credit_score,
debt_to_income_ratio, lender, interest_rate, loan_amount,
state, zip_code, loan_term, loan_to_value,
combined_loan_to_value, borrower_count, first_time_homebuyer,
loan_purpose, property_type, unit_count, occupancy_status,
product_type, co_borrower_credit_score, income)
recommendations.append({
'param':
'credit score',
'term':
'long',
'values':
value_array,
'probs':
prob_array,
'text':
'If your credit scrore increases to ' +
str(result_credit_score_high) +
", your chance of foreclosure will go down to " +
str(int(100 * prob_array[0])) + '%.'
})
result_down_payment_low, result_down_payment_high = find_optimum_down_payment(
foreclosure_model, foreclosure_scaler, borrower_credit_score,
debt_to_income_ratio, lender, interest_rate, loan_amount, state,
zip_code, loan_term, loan_to_value, combined_loan_to_value,
borrower_count, first_time_homebuyer, loan_purpose, property_type,
unit_count, occupancy_status, product_type,
co_borrower_credit_score, income, down_payment, 60)
print("prining optimum values for down payment")
print(result_down_payment_low)
print(result_down_payment_high)
if (result_down_payment_high != -1):
value_array, prob_array = create_array_for_chart(
'down_payment', down_payment, result_down_payment_high,
foreclosure_model, foreclosure_scaler, borrower_credit_score,
debt_to_income_ratio, lender, interest_rate, loan_amount,
state, zip_code, loan_term, loan_to_value,
combined_loan_to_value, borrower_count, first_time_homebuyer,
loan_purpose, property_type, unit_count, occupancy_status,
product_type, co_borrower_credit_score, income)
recommendations.append({
'param':
'down payment',
'term':
'short',
'values':
value_array,
'probs':
prob_array,
'text':
'If your down payment increases to ' +
str(result_down_payment_high) +
", your chance of foreclosure will go down to " +
str(int(100 * prob_array[0])) + '%.'
})
#salary
result_salary_low, result_salary_high = find_optimum_salary(
foreclosure_model, foreclosure_scaler, borrower_credit_score,
debt_to_income_ratio, lender, interest_rate, loan_amount, state,
zip_code, loan_term, loan_to_value, combined_loan_to_value,
borrower_count, first_time_homebuyer, loan_purpose, property_type,
unit_count, occupancy_status, product_type,
co_borrower_credit_score, income, income * 2)
print("prining optimum values for salary")
print(result_salary_low)
print(result_salary_high)
if (result_salary_high != -1):
value_array, prob_array = create_array_for_chart(
'salary', income, result_salary_high, foreclosure_model,
foreclosure_scaler, borrower_credit_score,
debt_to_income_ratio, lender, interest_rate, loan_amount,
state, zip_code, loan_term, loan_to_value,
combined_loan_to_value, borrower_count, first_time_homebuyer,
loan_purpose, property_type, unit_count, occupancy_status,
product_type, co_borrower_credit_score, income)
recommendations.append({
'param':
'salary',
'term':
'long',
'values':
value_array,
'probs':
prob_array,
'text':
'If your salary increases to ' + str(result_salary_high) +
", your chance of foreclosure will go down to " +
str(int(100 * prob_array[0])) + '%.'
})
find_optimum_bank(foreclosure_model, foreclosure_scaler,
borrower_credit_score, debt_to_income_ratio, lender,
interest_rate, loan_amount, state, zip_code,
loan_term, loan_to_value, combined_loan_to_value,
borrower_count, first_time_homebuyer, loan_purpose,
property_type, unit_count, occupancy_status,
product_type, co_borrower_credit_score)
#find zip codes
api_call_zip = "https://www.zipcodeapi.com/rest/QsRqONATNkmmVzzIt6wEhzs6d7Q2zjiQnSCXNGr25Y9FDlp8gvwHX4TAA40DMW7N/radius.json/" + request.form[
"zip"] + "/120/mile"
print(api_call_zip)
response = requests.get(api_call_zip)
#print(response.text)
if (len(response.text) > 0):
zip_dict = json.loads(response.text)
zip_dict['zip_codes'].sort(key=lambda x: x.get('distance'))
z_list = []
min_proba = int(foreclosure_probability * 100)
min_zip = -1
min_dist = -1
for z in zip_dict['zip_codes']:
zp = z['zip_code']
zp_short = int(zp[0:3])
st = categoricals.CATEGORICAL_INDEX_MAP['property_state'][
z['state']]
if (zp_short not in z_list):
z_list.append(zp_short)
foreclosure_probability_zip = predict_single_data(
foreclosure_model, foreclosure_scaler,
borrower_credit_score, debt_to_income_ratio, lender,
interest_rate, loan_amount, st, zp_short, loan_term,
loan_to_value, combined_loan_to_value, borrower_count,
first_time_homebuyer, loan_purpose, property_type,
unit_count, occupancy_status, product_type,
co_borrower_credit_score)
if (int(foreclosure_probability_zip * 100) < min_proba):
print('foreclosure_probability_zip: ' +
str(foreclosure_probability_zip))
print('zip: ' + str(zp_short))
print('zipcode: ' + str(zp))
print('distance: ' + str(z['distance']))
min_proba = int(foreclosure_probability_zip * 100)
min_zip = str(zp)
min_dist = z['distance']
if (min_proba <= 20):
break
print(z_list)
if (min_zip != -1):
recommendations.append({
'param':
'zip code',
'term':
'other',
'values': [],
'probs': [],
'text':
'If you are willing to buy a house in ' + str(min_zip) +
" instead (" + str(min_dist) +
" miles away from your initial selection), your chance of foreclosure will go down to "
+ str(min_proba) + '%.'
})
ranking = foreclosure_data_per_state[
request.form['property_state']]['ranking']
recommendations.append({
'param':
'other',
'term':
'other',
'values': [],
'probs': [],
'text':
'There are ' + str(ranking) +
' states with lower foreclosure rates than your selected state. If you are willing to move to a different state, you can use our <a href="/predict">prediction tool</a> to estimate your foreclosure probability in other states. You can also use our <a href="/vis">visualizations</a> to compare foreclosure rates and other demographics for different states.'
})
return render_template('predict_page.html',
pred=pred,
options=ALL_SELECTION_OTIONS,
selections=request.form,
recommendations=recommendations)
