Sectoral Analysis for Indian Economy

Top-Down Investment Strategy Optimization with Time Series Forecasting

Objectives: Model should be capable of capturing cyclical trends and patterns within different sectors of the Indian economy. Additionally, the model should generate lead and lag indicators to forecast the performance of specific sectors in future periods which would in turn enable one to make informed investment decisions based on the model outcome directly.

Why Nifty Index: To effectively evaluate the performance of different sectors, we decided to utilize the Nifty sectoral indices values as the target variable for different sectors. Nifty sectoral indices serve as reliable indicators to assess the performance, track trends, and gauge overall sentiment within specific sectors of the Indian stock market. This is because the index is a basket of stocks that represent the sector, and its performance is a reflection of the performance of the underlying stocks.

Now ,these are the 3 objectives that this project tries to fulfill-which are

1. Building forecasting models that effectively capture and predict business cyclical shifts within specific industries.
2. Creating an automated tool that can compute lead/lag signals, enabling real-time identification of trends and patterns.
3. Enhancing the decision-making process by providing accurate forecasted trends.

Project Pipeline

1. Data Collection and Preprocessing:

• Identify potential features from sectoral annual reports, research papers, government websites, etc.
• Collect relevant data from sources like CMIE & World Bank.
• Import data into Jupyter Notebook.
• Trim spaces, remove commas, and set appropriate indices.
• Handle missing values and outliers.

2. Exploratory Data Analysis (EDA):

• Generate summary statistics for the dataset.
• Create distribution plots and line plots for a deeper understanding.
• Explore relationships between variables using line plots.

3. Feature Selection and Engineering:

• Perform correlation analysis to identify high correlations.
• Drop features with high correlation coefficients to reduce multicollinearity.
• Use Granger causality test to determine causal relationships.
• Adjust the dataset based on lag order from Granger causality.

4. Model Building using Multiple Linear Regression (MLR):

• Split data into training and testing sets.
• Apply scaling transformations to predictor variables.
• Check MLR assumptions (linearity, independence, homoscedasticity, normality).
• Fit MLR model using stepwise backward elimination.
• Validate model using appropriate evaluation metrics.
• Compare actual vs. predicted values.

5. Time Series Modeling using SARIMAX:

• Create line plots to visualize time series data.
• Perform time series decomposition.
• Check stationarity visually and using Dickey-Fuller test.
• Apply differencing for stationarity.
• Choose optimal orders for SARIMA model based on AIC, BIC, and HQIC.
• Fit and validate SARIMA model.

6. Iterative Improvement (Dynamic Model):

• Gather feedback from stakeholders and domain experts.
• Identify areas for improvement based on feedback.
• Update pipeline and models iteratively.
• Continuously incorporate new data for model refinement.

7. Forecasting and Model Selection:

• Make forecasts using the validated models.
• Compare the performance of different models.
• Choose the most appropriate model based on evaluation metrics.
• Generate forecasts for risk-adjusted returns in the investment strategy.

Remember, this pipeline is a general outline of the steps you can follow in your project. Adjustments might be needed based on specific data characteristics and project requirements.