Time-Series-Modeling: Hypixel Skyblock Cocoa Market

A production-ready data engineering and modeling pipeline designed to ingest, sanitize, and reconstruct high-velocity time-series data from the Hypixel Skyblock Cocoa Bean market. This framework addresses critical data quality issues by employing multivariate, model-based imputation techniques that maintain strict temporal causality.

Pipeline Architecture

Project Context

The Hypixel Skyblock economy is a highly volatile, player-driven marketplace. Accurate price discovery is difficult due to sparse data and missing order-book entries. This project serves as a robust quantitative engine to:

• Prevent Look-Ahead Bias: Implements chronological data processing to ensure forecasting models are trained strictly on past observations.
• Recover Missing Market Intelligence: Uses a sequential XGBoost imputation pipeline to reconstruct fragmented minBuy and minSell price points, the very edge of the order book.
• Bridge Market Data: Leverages autoregressive (AR) feature engineering and rolling window statistics to provide a continuous, model-ready panel for arbitrage analysis.

Repository Structure

Time-Series-Modeling/
├── analysis/
│   └── modeling.ipynb       # Jupyter notebook for research, backtesting, and visualization
├── data/
│   ├── raw/                 # Unprocessed API response
│   └── cleaned/             # Imputed panel ready for modeling
├── scripts/
│   ├── data_construction.py # Data ingestion and feature engineering
│   ├── market_api.py        # Logic for interacting with Hypixel marketplace endpoints
│   └── ts_imputer.py        # Custom multivariate XGBoost imputation engine
└── .gitignore               # Standard Python VSCode GitIgnore

Technical Approach

The pipeline follows a rigorous order of operations to maintain statistical integrity:

1. Temporal Alignment: Raw timestamps are parsed and sorted chronologically to eliminate data leakage.
2. Dense Interpolation: Highly continuous variables (e.g., buy, sell, volumes) are bridged using linear interpolation.
3. Temporal Feature Engineering: AR(1) spot lags and moving weekly window features are injected to capture momentum.
4. Sequential Imputation: XGBoost is deployed to surgically reconstruct the sparse boundary layers (minBuy/minSell) of the market, ensuring the imputed values are statistically consistent with observed global market trends.

Getting Started

This repository is built for modularity. To integrate the construction pipeline into your own analysis, ensure the project root is in your sys.path:

import sys
sys.path.append('path/to/Time-Series-Modeling')

from scripts.data_construction import data_construction

Returns a clean, model-ready panel

df = data_construction(path="data/raw/cocoa_beans_historical.csv")