KSH Inflation Analysis

This repository contains a small, modular Python project for processing annual CPI (consumer price index) data exported from the Hungarian Central Statistical Office (KSH). The code reads a KSH CSV, computes annual inflation as CPI_index - 100, produces diagnostic plots, and fits a simple linear regression of inflation on year.

Contents

• main.py — the top-level orchestrator: parses CLI args, loads data, prepares it, runs regression, saves outputs and produces plots.
• inflacio/ — package with small modules: io.py, prep.py, analysis.py, plots.py.
• stadat-ara0001-1.1.1.1-hu.csv — example KSH CSV (included in repository).
• requirements.txt — Python dependencies.
• README.md — original README (Hungarian).
• README_en.md — this English README.
• output/ — example output files produced by main.py (cleaned CSV, regression metrics, plots).

Requirements

The project depends on these Python packages (also listed in requirements.txt):

• pandas
• numpy
• matplotlib
• scikit-learn
• statsmodels

Install dependencies (PowerShell):

python -m pip install -r .\requirements.txt

If python is not available on your PATH, use your system Python executable instead.

Usage

Run the analysis from the project root:

python .\main.py --input stadat-ara0001-1.1.1.1-hu.csv --outdir output

Options:

• --input, -i : path to the input CSV (optional — the script also tries a couple of default filenames).
• --outdir, -o : output directory (default: output).
• --no-show : do not display plots interactively (useful for headless runs).
• --debug : enable debug logging.

What the script does:

• Loads the KSH CSV (tries several encodings; expects ; separator and , decimal mark).
• Detects the CPI column and the year column, converts types and drops missing rows.
• Computes inflation percentage as: Inflacio_pct = CPI_index - 100.
• Saves cleaned data to output/cleaned_data.csv.
• Fits a linear regression of inflation(%) on centered year (sklearn + statsmodels), computes R² and RMSE.
• Saves OLS summary to output/regression_summary.txt and a compact JSON metrics file to output/regression_metrics.json.
• Produces and saves plots to output/plots/ (line plot, scatter + regression, residuals, residuals histogram).

Output files

• output/cleaned_data.csv — prepared dataset used in the analysis.
• output/regression_summary.txt — full statsmodels OLS summary.
• output/regression_metrics.json — JSON with slope, intercept, R², RMSE and slope p-value.
• output/plots/*.png — generated plot images.

Interpretation (short)

• CPI_index is the annual index (100 = previous year). For example, 104.9 indicates ~4.9% increase.
• Inflation (%) = CPI_index - 100.
• Linear regression fits a trend line: Inflation(%) = slope * Year + intercept.
  • R² measures how much variance the linear model explains.
  • RMSE is the average error in percentage points.
  • slope p-value indicates whether the trend is statistically significant.
• Inspect residual plots: if residuals show patterns, a simple linear trend may be insufficient.

Development & Improvements

• Allow explicit CLI selection of the year and CPI columns if automatic detection fails.
• Add unit tests for load_ksh_csv, detect_cpi_column, prepare_dataframe, and run_regression.
• Add confidence intervals or prediction bands to the regression plot.
• Add time series models (ARIMA/Holt-Winters) for forecasting.

Troubleshooting

• If the script cannot find the CSV, pass it with --input or place it in the project root.
• If plotting fails on headless servers, use --no-show and check the output/plots/ folder for saved PNGs.
• If dependencies are missing, re-run pip install -r requirements.txt.