bandicoot: a Python Toolbox for Mobile Phone Metadata

Authors: Yves-Alexandre de Montjoye, Luc Rocher, Alex Sandy Pentland

JMLR 2016 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We tested bandicoot on a computer with an Intel i7 CPU (2.6GHz) and 8GB of memory for users with an average of 20 records per days over 3 months. It takes on average 250ms to compute all 1442 behavioral and mobility indicators using the standard Python interpreter, CPython, and 160ms using pypy, a fast just-in-time compiler. For all indicators, including network features, the total time is 1.11s using CPython (740ms with pypy).
Researcher Affiliation Academia Yves-Alexandre de Montjoye de EMAIL MIT Media Lab, 02139 Cambridge MA, USA Imperial College London, Dept. of Computing and Data Science Institute, London SW7 2AZ, UK Luc Rocher EMAIL Universit e catholique de Louvain, ICTEAM, 1348 Louvain-la-Neuve, Belgium Alex Sandy Pentland EMAIL MIT Media Lab, Cambridge, 02139 Cambridge MA, USA
Pseudocode No The paper describes the functionalities and performance of the bandicoot toolbox but does not include any structured pseudocode or algorithm blocks.
Open Source Code Yes bandicoot is an open-source Python toolbox to extract more than 1442 features from standard mobile phone metadata. ... bandicoot is hosted on Git Hub and has already been developed by 9 contributors over the last three years...
Open Datasets No The paper discusses mobile phone metadata in general and mentions testing bandicoot with users having 'an average of 20 records per days over 3 months', but it does not specify a publicly available dataset with concrete access information (link, DOI, citation) that was used for its experiments.
Dataset Splits No The paper describes testing bandicoot's performance using data for 'users with an average of 20 records per days over 3 months', but it does not specify any dataset splits (e.g., training, validation, test sets, or percentages).
Hardware Specification Yes We tested bandicoot on a computer with an Intel i7 CPU (2.6GHz) and 8GB of memory for users with an average of 20 records per days over 3 months.
Software Dependencies No bandicoot runs on Python 2 and 3... To make its use (incl. in Hadoop environments) and installation easier, we developed bandicoot to be free of any dependencies such as pandas or compilers.
Experiment Setup Yes We tested bandicoot on a computer with an Intel i7 CPU (2.6GHz) and 8GB of memory for users with an average of 20 records per days over 3 months. It takes on average 250ms to compute all 1442 behavioral and mobility indicators using the standard Python interpreter, CPython, and 160ms using pypy, a fast just-in-time compiler. For all indicators, including network features, the total time is 1.11s using CPython (740ms with pypy).