Skip to content

Latest commit

 

History

History
65 lines (41 loc) · 3.19 KB

report.md

File metadata and controls

65 lines (41 loc) · 3.19 KB
Raw

Economic Recession - Report

Overview of the Analysis

  • The goal is to create a Machine Learning model using historcial Treasury rates and Gross Domestic Product(GDP) data to predict US recession.

  • Data Source -

    1. 10 Years Treasury Historical Interest Rates since 1959 (Source : FRED API - Federal Reserve Bank of St.Louis)

    2. 2 Years Treasury Historical Interest Rates since 1974 (Source : FRED API - Federal Reserve Bank of St.Louis)

    3. 1 Year Treasury Historical Interest Rates since 1959 (Source : FRED API - Federal Reserve Bank of St.Louis)

    4. Gross Domestic Product (GDP) Data since 1959 (Source: BEA - US Bureau of Economic Analysis)

  • Following data points are needed to build this model -

    • 10 Years Minus 2 Years Historical Treasury Interest Rates
    • 10 Years Minus 1 Year Historical Treasury Interest Rates
    • Change in private inventories

  • The target feature 'IS_RECESSION' is caculated using GDP quaterly growth rates. IS_RECESSION is represented as 'True' when GDP Growth rate is negative, otherwise it is recorded as 'False'.

  • The data downloaded from files and API are loaded into Pandas DataFrames.

  • The data was then split into training and test set.

    • The target column 'IS_RECESSION' was pulled into a seried named 'y'
    • All the other features were pulled into a dataframe named 'X'
    • Scikit-Learn 'train_test_split' function was used to split X and y to train and test data sets named X_train, y_train and X_test, y_test.

  • Logistic Regression model from Scikit Learn was chosen to train and predict the recession

    • LogisticRegression model was created using solver named 'lbfgs' and random_state=42, test_size=.2 and stratify=y
  • LogisticRegression Model created was then fitted using training dataset - X_train and y_train.
  • The model is then used to predict recession of test dataset 'X_test'

Model Optimizations

The accuracy score was recorded as 95.38% for the original model. The precision and recall for 'Is_Recession' label 'True' was 0.

To improve accuracy, precision, and recall, the following optimizations were done-

  • An additional feature "Change in Private Inventories' was introduced.
  • For the train and test split, the test size was adjusted to 20%, and stratify was set to 'y'

Following ML algorithms were tried out but didn't improve the score and hence wasn't considered for this model.

  • K-Nearest Neighbors (KNN) algorithm
  • Decision Tree
  • XGBoost

Results

  • Logistic Regression Model :
    • Classification Report
      • Accuracy (0.98),
      • IS_RECESSION False (No Recession) - Precision (0.98), and Recall (1.00).
      • IS_RECESSION True (Recession) - Precision (1.00), and Recall (0.50).

Summary

  • The optimized version of the Logistic regression model for predicting an economic recession based on differences in the interest rates of long-term and short-term treasuries and changes in private inventories achieved an accuracy score of 98.07%.
  • The precision and recall for ‘Is_Recession’ label ‘False’ are .98 and 1.00 respectively. The precision and recall for the ‘Is_Recession’ label ‘True’ are 1.0 and .50 respectively

Recommends Logistics Regression Model for predicting an economic recession