Project Structure

├── data_clean
├── data_raw
├── documentation
├── results
│   ├── Plots
├── scripts
└── sql

Data Extraction

All the data were downloaded from respective sources manually, by selecting indicators, fields or specific files. The downloaded raw data were stored in data_raw directory. The data ranges from minimum 1980 to maximum 2024.

• BD_economic_indicators.csv has the Kaggle dataset. It contains Macroeconomic indicators of bangladesh.
• world_bank1.csv has a World Bank dataset. It contains some Macroeconomic and Banking indicators.
• world_bank2.csv has another World Bank dataset. It also contains some Macroeconomic and Banking indicators.
• IMF-FSI-A.BD.FSGDLD_XDC.csv has a small IMF-FSI dataset containing Total loans in millions data specificly.
• IMF-FSI-A.BD.FSKNL_PT.csv has IMF-FSI data of Net non-performing loans in percentage.
• IMF-FSI-A.BD.FSANL_PT.csv has IMF-FSI data of Non-performing loans of total loans in percentage.
• IMF-FSI-A.BD.FSCD_PT.csv has IMF-FSI data of Customer deposits to total loans in percentage.

Cleaning Data

The raw data were cleaned using python scripts in scripts directory and saved in data_clean directory.

• BD_macroeconomic_clean.py, world_bank1_clean.py, world_bank2_clean.py, GDLD_imf_clean.py, KNL_imf_clean.py, ANL_imf_clean.py, CD_imf_clean.py \(\Rightarrow\) Script names correspond directly to the files they process.
• clean_duplicate_values.py, clean_partition_sqlMacroData.py were tasked to clean some intermediate dataframes.
• recalculate_change.py, reshape_change.py were for modifying some dataframes to load in SQL database finally for analysis.

SQL

Postgresql was used for creating the database of the cleaned data along with the schema, doing test analysis, saving data tables and exporting shaped tables for visualization. In the sql directory files –

• schema.sql defined tables dim_time, dim_indicator, dim_source, fact_macro, fact_banking. The latter 2 were the tables containing the suggestive indicator values while the prior ones were purely for mapping in between the latter without any ambiguity.
• load_macro.sql, load_fsi.sql, load_wb.sql were suggestively to load the cleaned data from data_clean directory.
• views.sql defined tables joining the original tables to make the analysis easier.

The analytical queries were done in the interactive database sessions in local system terminal and the necessary ones were saved as csv extracts for visualization in the results directory.

Results

The following results were used as inputs for the analysis presented in subsequent sections.

Macroeconomic Results

The macroeconomic indicator sub-tables as results are:

• gdp_per_capita_current_us.csv
• inflation_consumer_prices_annual.csv
• inflation_rate.csv
• total_investment.csv
• government_debt.csv
• unemployed_rate.csv

Banking Results

The banking sector indicator sub-tables as results are:

• bank_nonperforming_loans_to_total_gross_loans.csv
• deposits_to_loans_pct.csv
• loans_to_domestic_economy_million_BDT.csv
• loans_total.csv

Cross-linked Results

The results that are indicative of the cross-sector outcomes:

• broad_money_of_gdp.csv
• domestic_credit_to_private_sector_of_gdp.csv