Time-Based Library Borrowing Transaction Dataset for Predictive Modeling and Temporal Pattern Analysis (2018–2025)

This dataset contains anonymized borrowing transaction records from a university library, enriched with temporal features to support predictive modeling and analysis of borrowing patterns.

The dataset consists of 52,394 transaction records collected between 2018 and 2025, where each record represents an individual borrowing event. It was developed to support the study “Time-Based Library Borrowing Patterns: A Predictive Analysis using Random Forest Regression”.

Each transaction includes the following attributes:

borrow_date: Date of the borrowing transaction

user_id: Anonymized user identifier

book_id: Unique identifier of the borrowed item

exam_period: Indicator of academic examination period (e.g., 0 = non-exam, 1 = exam period)

hour_of_day: Hour when the borrowing occurred (0–23)

day_of_week: Day of the week (e.g., Monday–Sunday)

month: Month of the transaction (1–12)

This dataset enables:

Predictive modeling of borrowing transaction volumes

Temporal analysis of user borrowing behavior

Feature importance analysis in machine learning models

Benchmarking regression and time-aware algorithms

In the associated research, a Random Forest Regression model trained on this dataset achieved strong predictive performance (R² = 0.867, MAE = 10.89, RMSE = 14.91), with hour_of_day identified as the most influential feature.

All user-related data have been anonymized to ensure privacy and ethical data sharing. This dataset is suitable for research in machine learning, temporal data mining, demand forecasting, and smart library systems.