Вы находитесь на странице: 1из 16

FINANCE & RISK ANALYTICS

PROJECT REPORT

Sharwani Das
06-SEP-2020
PART 1 CREDIT RISK:
Read the data as an appropriate Time Series data and plot the data.

No Duplicate Values:

Data Summary:

1
Outlier Treatment:
Check the Outliers with below method and Dropping records higher than upper whisker & Capping values lower than lower whisker:

Missing Value Treatment:

List of columns numbers having missing value:

Impute the missing values with ‘Median’, using


‘SimpleImputer’.

New Variable Creation:


Three new variables have been created – Profit Margin, Return On Total Assets & Debt To Equity Ratio

2
Also, the target variable ‘Default’ has been derived from ‘Networth_Next_Year’.
Below variables have been delated after creating new variables -
'Company_Code'
'Networth_Next_Year'
'PBT','Total_Liabilities'
'Total_Equity'
'Total_Sales'
'total_assets'

Univariate & Bivariate Analysis:

3
4
5
6
Pair Plot:

7
Heat Map:

Net Working Capital vs Default:

8
Net Working Capital Vs Net Worth Next Year

PBT Vs Net Worth Next Year

Profit Margin Vs PBT To Total Income

9
Split into Test & Train Data:

Columns after splitting:

MODELING:
Most important variables have been identified after multiple iterations of variable combination and fitting in StatsModels Logit
function. Below variables has been identified after 12 iterations:

The ‘Default’ variable ratio is 0.127, after checking different ratio the Cut-off value is identified as 0.15:

10
Train & Test Data – Performance Metrics

TRAIN DATA TEST DATA

Altzman Score - Train & Test Data:

The Altman’s Z-score formula : ζ = 1.2A + 1.4B + 3.3C + 0.6D + .999E Zeta (ζ) is the Altman’s Z-score

● A is the Working Capital/Total Assets ratio


● B is the Retained Earnings/Total Assets ratio
● C is the Earnings Before Interest and Tax/Total Assets ratio
● D is the Market Value of Equity/Total Liabilities ratio
● E is the Total Sales/Total Assets ratio

11
Alt_train['Altman_score']= 1.2*(Alt_train.Net_Working_Capital/Alt_train.total_assets) +
1.4*(Alt_train.Retained_Earnings/Alt_train.total_assets) +
3.3*(Alt_train.PBIT/Alt_train.total_assets) +
0.6*(Alt_train.MarketValue_Equity/Alt_train.Total_Liabilities) +
0.999*(Alt_train.Total_Sales/Alt_train.total_assets)

The model has been evaluated using a standard threshold of 1.8.

TRAIN DATA TEST DATA

As per the Altman Score, the accuracy for Train & Test data is 67%, whereas Regression Model does much better prediction with
Training data accuracy – 84% & Test Data accuracy – 85%.

Even the Sensitivity & Specificity is also better for the Logistic Regression model.

12
PART 2 – MARKET RISK:

Market Risk Data set:

13
Draw Stock Price Chart for any 2 variables:

Year Vs SAIL Year Vs Shree Cement

Calculate Returns:

Create the Return Data set after dropping ‘Date’ Column:

14
Calculate Stock Means and Standard Deviation
• Stock Means: Average returns that the stock is making on a week to week basis
• Stock Standard Deviation: It is a measure of volatility meaning the more a stock's returns vary from the stock's average
return, the more volatile the stock

STOCK MEAN STOCK STD DEVIATION

Stock with the lowest mean has the highest standard deviation, and as the Stock Mean increases Standard Deviation decreases It
plays a major role in a portfolio that has competing stock with more returns & less risk.

• One with highest return and lowest risk &


• One with lowest risk and highest return

Therefore from pure Returns perspective, Shree_Cement looks good in this dataset & from pure Risk perspective (as measured by
standard deviation), Infosys followed by Shree_Cement looks good in this dataset.

15

Вам также может понравиться