https://packt.link/EarlyAccessCommunity
In this chapter we will analyze data from an early Kaggle competition, Data Science for Good: Kiva Crowdfunding (see reference [1]). You will learn how to tell a story with data in a way that is both informative and catching for the reader. Then we will verify through a detailed analysis a hypothesis from another competition dataset, about Kaggle Metadata.
Approaching a new dataset resembles sometime to an archeological excavation and sometime to a police investigation. We proceed to unearth hidden valorous insights from under a pile of data or we try to uncover elusive evidence by a systematic, sometime arride process, that resemble either the technical discipline of the archeologist or the method of the detective. All data can tell a story. It is the analyst choice if this story is told in the style of a scientific report or in the vivid, attractive form of a detective novel. In this chapter, we will combine techniques we developed in previous chapters for analysis of tabular- numerical and categorical, text and geospatial data and combining data from multiple sources and show how to tell a story with data.
Kiva.org is an online crowdfunding platform that has the mission to extend to the poor and financially excluded people around the world the benefits of financial services. Those people can benefit, through services of Kiva, to borrow small amounts of money. These microloans are provided by Kiva through their partnerships with financial services institutions in the countries where resides the receiver of the loans. In the past, Kiva has provided in the targeted communities, over one billion US dollars in microloans. To extend the reach of their assistance, and at the same time, to improve the understanding of specific needs, factors that make the whole difference for impoverished people in different parts of the world, Kiva wanted to better understand the particular conditions of each potential borrower. Due to the diversity of the problems in different parts of the world, the specificity of each case, the multitude of influencing factors, the mission of Kiva to identify...
The competition requires the participants to identify or collect relevant data, besides the data provided by organizers. This includes loans information, Kiva global multidimensional poverty index (MPI) by region and location, loan theme and loan themes by region. The loans information include an unique id, loan theme id, loan theme type, local financial organization partner id, funded amount (how much Kiva provided to the local partner), loan amount (how much the local partner disbursed to the borrower), the activity of the borrower, sector, use of the loan, country code, country name, region, currency, posted time, disbursed time, funded time, duration in time for which the loan was disbursed, the total number of lenders that contributed to a loan, gender of borrowers, and repayment interval. Kiva MPI by region and location include region or country name, ISO-3 code for the country, region, world region, MPI value and the geolocation (latitude and longitude) of the current...
It is important to stress from the start that one good solution for an analytics competition is not necessarily a complete exploratory data analysis. From experience with several analytics competitions and looking to the highest ranked solutions, sometime is quite the opposite. Criteria for scoring a solution of an analytics competition are changing in time, but some are repeatedly adopted. The evaluators will frequently look to the originality of the approach, to the composition and to the documentation. To obtain high scores on all these criteria, the authors will have to prepare very well. An extended exploration of the data is still necessary, so that all results presented can be well documented. While useful for the research, this approach doesn’t need to be fully included in the narrative of the solution Notebook. Actually, the author could select and discuss, in his story, a small part of the data, as long as the narrative...
© 2023 Packt Publishing Limited All Rights Reserved
