Aayush Tiwari, home
Résumé
§2 Selected work

Indian IPO Analytics

NSE/BSE market data analysis and risk clustering

Status
Reproducible exploratory analysis
Role
Sole public repository contributor; built the notebook analysis, generated figures, and local Dash application.
Domain
Data Analytics / Applied Statistics
Source
Repository (opens in a new tab)
Figure 1Shows the analysis path over the curated dataset alongside the project's own results: a 0.9808 five-fold cross-validated R², a median absolute error near 4.15%, and two K-Means risk clusters across 63 records in 27 sectors. These describe that dataset only and are not prospective.
§1Brief

The problem this had to solve.

Turn a small curated Indian IPO dataset into an inspectable analysis of listing performance, demand, issue size, sector, and post-listing returns without presenting the results as a trading model.

Indian IPO Analytics studies 63 NSE/BSE IPO records across 27 sectors from 2019 through 2024. The project builds a 34-column table with complete enriched return coverage for its dataset.

Regression, clustering, and distribution analysis feed an interactive Plotly Dash dashboard. Results include approximately 0.9808 five-fold cross-validation R², approximately 4.15% MAE, and two K-Means risk clusters; these remain project-dataset findings rather than general financial predictions.

Constraints

  1. The dataset contains 63 IPOs across 27 sectors from April 2019 through November 2024.
  2. Grey Market Premium is both an input and a close market expectation of listing performance, so very high listing-gain regression scores require cautious interpretation.
  3. Post-listing return enrichment uses cached/fallback values as well as yfinance; complete coverage does not mean every value was fetched live from one source.
  4. The Dash application is documented for local use at localhost and no maintained public deployment is listed.
§2Architecture

How the system is put together.

NSE/BSE IPO records

63 records across 27 sectors from 2019–2024

source

Data enrichment

34-column table with complete project-dataset return coverage

process

Statistical analysis

Regression, clustering, and distributions

process

Risk clusters

Two K-Means groups in the project analysis

decision

Dash dashboard

Interactive Plotly exploration

output
§3Decisions

What was chosen, and what it cost.

Keep the workflow inspectable

Choice
Ship the notebook, generated figures, cached return data, and an interactive Dash application in the same repository.
Rationale
Readers can review both the analysis steps and the presentation layer rather than seeing only headline results.

Use several analytical lenses

Choice
Combine exploratory plots, regression, sampling, K-Means clustering, distribution fitting, and time summaries.
Rationale
The dataset supports descriptive questions beyond a single prediction score.
Cost
Many reported findings come from the same small curated sample and are not independent validation studies.

Qualify model results as dataset findings

Choice
Report held-out and five-fold regression results while keeping the portfolio language scoped to the 2019–2024 dataset.
Rationale
The model has not been evaluated prospectively or against a later market period.
§4Validation

How it was checked.

  • The repository includes the executed notebook, eight generated figure files, and a local Plotly Dash application.
  • The documented regression run reports train R² 0.9922, test R² 0.9834, test MAE 4.15%, and five-fold CV R² 0.9808 ± 0.0117.
  • The clustering study selects k=2 with silhouette score 0.4740, while distribution checks reject a Normal fit at the 5% level for this sample.
Indian IPO Analytics overview combining exploratory plots from the project dataset.
Figure 2 · Repository artifactGenerated exploratory analysis from the repository's curated 63-record, 2019–2024 IPO dataset; the findings are descriptive rather than prospective forecasts. Inspect the source (opens in a new tab)
§5Evidence

What was measured.

Each figure states the conditions it was measured under. None of them describe production traffic.

63

IPO records

NSE/BSE IPOs in the 2019–2024 project dataset.

27

sectors

Sector coverage in the enriched analytics table.

0.9808

5-fold CV R²

A result of this project dataset, not a universal prediction claim.

~4.15%

MAE

A project-dataset regression result.

§6Limits

Where this stops being true.

  • The sample is small, curated, and limited to one country and market period; the findings are not a prospective return forecast.
  • The repository does not document a leakage audit, temporal validation split, uncertainty intervals for predictions, or transaction-cost analysis.
  • GMP alone explains 99.0% of listing-gain variance in the documented run, which limits what the regression demonstrates about less direct features.
  • No automated test suite or hosted dashboard is present in the public repository.
§7Technology

What it is built with.

Runtime & interface

  • Python
  • Dash

Data & evaluation

  • Pandas
  • NumPy
  • SciPy
  • Scikit-learn
  • Statsmodels
  • yfinance

Infrastructure & providers

  • Plotly
Explore the repository (opens in a new tab)
Next documentCONCLAVE