IME - Forecasting Macroeconomic Factors for Enterprise Risk Management

mohamed-khamis
mohamed-khamis Partner, L2 Admin, L2 Designer, Dataiku DSS Core Designer, Dataiku DSS & SQL, Dataiku DSS ML Practitioner, Dataiku DSS Core Concepts, Snowflake Advanced, Dataiku DSS Adv Designer, Registered, Dataiku DSS Developer, Dataiku Frontrunner Awards 2021 Participant, Frontrunner 2022 Participant Posts: 13 Partner

Name: Mohamed AbdElAziz Khamis Omar

Title: Head of Data Science

Country: Egypt

Organization: IME

IME is a key market player in Data Management solutions. To learn more: www.infme.com

Awards Categories:

  • Data Science for Good
  • Moonshot Pioneer(s)

Business Challenge:

  • Forecasting macroeconomic factors for enterprise risk management solution.
  • The bank needs to forecast macroeconomic factors as input for Probability of Default and Expected Credit Loss modules.

Business Solution:

Build a time series forecasting model using Dataiku.

File Source:

image38.png

Examples on the selected Macroeconomic variables

  1. Unemployment percentage is the percentage of the total labor force that is unemployed but actively seeking employment and willing to work.
  2. GDP Gross domestic product is the monetary value of all finished goods and services made within a country during a specific period. GDP provides an economic snapshot of a country, used to estimate the size of an economy and growth rate. GDP can be calculated in three ways, using expenditures, production, or incomes.
  3. Consumer Price Index (CPI) is a price index, the price of a weighted average market basket of consumer goods and services purchased by households. Changes in measured CPI track changes in prices over time.
  4. Exchange rate is the rate at which one currency will be exchanged for another currency. Inflation is the rate of increase in prices over a given period of time.
  5. Inflation is typically a broad measure, such as the overall increase in prices or the increase in the cost of living in a country.
  6. Population typically refers to the number of people in a single area, whether it be a city or town, region, country, continent, or the world.
  7. Interest rate is the amount of interest due per period, as a proportion of the amount lent, deposited, or borrowed (called the principal sum).
  8. Gas prices may refer to Gasoline and diesel usage and pricing.
  9. Stock Market, equity market, or share market is the aggregation of buyers and sellers of stocks (also called shares), which represent ownership claims on businesses; these may include securities listed on a public stock exchange, and stock only traded privately, such as shares of private companies which are sold to investors through equity crowdfunding platforms.

image40.pngimage1.png

image39.png

image42.png

image41.png

image12.png

Training dataset preparation

image51.png

Design

General settings

image49.png

Train/test set for final evaluation

image56.png

External Features

We are not using any external features due to somehow low correlation between GDP and other macroeconomic factors, e.g., inflation, employment rate, population, exchange rate.

image54.png

Examples on the utilized algorithms for macroeconomic variables are:

  • Seasonal trend first subtracts the data seasonality estimated with STL (Seasonal and Trend decomposition using Loess LOESS (locally estimated scatterplot smoothing)). It then forecasts deseasonalized data using an Exponential Smoothing model adjusted for trends.
  • Non-Parametric Time Series (NPTS) predictor predicts future value distribution by sampling from past observations with weights exponentially decreasing over time.
  • Simple Feed Forward is a simple neural network that forecasts probability distributions for the next forecast horizon values, given the preceding context length values.
  • DeepAR is an autoregressive recurrent neural network that forecasts probability distributions for the next forecast horizon values given the preceding context length values. It also uses lagged values and time features automatically computed based on the selected time-frequency.
  • The Mean Absolute Percentage Error, also known as mean absolute percentage deviation, is a measure of prediction accuracy of a forecasting method in statistics
  • Setting Forecast Horizon: 4 Years

Business Area: Accounting/Finance

Use Case Stage: Proof of Concept

Value Generated:

  • Forecasting Macroeconomic factors as an input for a Probability of Default prediction module; the bigger system is a Risk Management Module.
  • MAPE = 1.9% using the Seasonal Trend Macroeconomic Forecasting.

image29.png

image32.png

image33.png

image34.png

image35.pngimage36.png

Output from Evaluation

image37.png

image12.png

Output from Scoring

image14.pngimage15.png

Value Brought by Dataiku:

Built-in Time series forecasting.

Value Type:

  • Reduce cost
  • Reduce risk
  • Save time
Setup Info
    Tags
      Help me…