Researcher Influence Prediction (ResIP) — Dataset

This repository contains the dataset used to train and evaluate the models proposed in:

Researcher influence prediction (ResIP) using academic genealogy network Dhananjay Kumar, Plaban Kumar Bhowmick, Jiaul H. Paik Journal of Informetrics, Volume 17, Issue 2, Article 101392 (2023) https://doi.org/10.1016/j.joi.2023.101392

Overview

The dataset supports the Researcher Influence Prediction (ResIP) task — predicting a researcher's future influence in an academic field by modelling the growth of their academic lineage (family graph/tree) over time. It was built from a subset of researchers drawn from the Mathematics Genealogy Project (MGP).

Getting Started

Unzip the dataset archive:

tar -xvjf dataset.tar.bz2

The extracted folder contains multiple NumPy and pickle files, one set per method proposed in the paper.

File Naming

File names indicate which method/dataset they correspond to:

Method	Description
ResIP-M1	Single-point output prediction
ResIP-M2	Output sequence prediction
ResIP-M3	Identified by "ResIP-M3" in the file name

Data Format

ResIP-M1 and ResIP-M2

The train and test data for these methods are stored as tuples containing:

• Input — sequence of embedding vectors for the researchers in the family graph/tree
• Output:
  • ResIP-M1 — a single output value (single time-point prediction)
  • ResIP-M2 — an output sequence (multi time-point prediction)

ResIP-M3

Data files are identified by the "ResIP-M3" substring in their file name.

Common Notes

• Input: A sequence of embedding vectors representing the researchers in the family graph/tree at a given time point.
• Output: The family size during the prediction period.
• Scaling: Output values are not pre-scaled. Apply a log2 transformation to the output before use.

Loading the Data

Example of loading a NumPy or pickle file:

import numpy as np
import pickle

# NumPy files
data = np.load("path/to/file.npy", allow_pickle=True)

# Pickle files
with open("path/to/file.pkl", "rb") as f:
    data = pickle.load(f)

Remember to apply the log2 transformation to output values before scoring or evaluation:

transformed_output = np.log2(output)

Citation

If you use this dataset, please cite:

@article{kumar2023resip,
  title   = {Researcher influence prediction ({ResIP}) using academic genealogy network},
  author  = {Kumar, Dhananjay and Bhowmick, Plaban Kumar and Paik, Jiaul H.},
  journal = {Journal of Informetrics},
  volume  = {17},
  number  = {2},
  pages   = {101392},
  year    = {2023},
  doi     = {10.1016/j.joi.2023.101392}
}