On 27-28 May, ISET organized a Data Hackathon in the framework of the “Building Capacity in Modern Data Analysis in Georgia”, a project being carried out with Tartu University.

The attending students grouped into four teams, each of which discussed different problems and tried to find different solutions to a specific issue. Overall, the hackathon was an interesting practice experience with plenty of learning opportunities.

Each team was led by two students from the MDA Concentration. The participants were provided data to work with, and also benefited from the help of mentors who navigated group works.

The datasets were as follows:

1. Real estate market data (sale prices and rents in Tbilisi)

2. The locations and average prices of Tbilisi Restaurants scraped from www.tripadviser.com

3. SPAR daily sales data for two days (Wednesday and Sunday)

4. The time and locations SPAR branches were opened in Tbilisi

5. Hourly sales of SPAR Buffet for three-months span.

Alpha (Winning group)

This team worked with sale and rental prices of apartments in Tbilisi. They aimed to understand how sale prices and rent rates are distributed within different areas of the city, as well as what kind of factors affect the prices and demonstrate the main features of price determination; they also discussed investment possibilities in the Tbilisi real estate market. For this purpose, they exploited Random Forest, Propensity Score Matching (PSM) and K-Nearest neighbors (KNN) models in order to tackle classification problems. As a conclusion, they found that return time on investment is longer in the central districts. In addition, they showed that sale price is the most important determinant of return, and area is the most important feature explaining sale price.

Cab-Hub

This team worked with two datasets: Trip Advisor and Real Estate Datasets. For the Trip Advisor data, they conducted a geospatial analysis illustrating the density of Georgian restaurants in Tbilisi by their locations and city regions.

For the Real Estate dataset they conducted analysis based on the K-Means and Boosting algorithms with the goal to understand what the main factors are which affect the prices of houses for rent and sale in Tbilisi. A clustering scheme was used for this analysis, dividing the whole dataset in three main clusters based on the similar characteristics of houses, such as the number of rooms, bedrooms, balcony, and total area. Using the skills acquired thanks to the Geospatial Data Analysis course, this team managed to illustrate all their results by maps and geo-specifications.

Data

This group conducted a basket analysis using two days’ worth of Spar sales for more than four thousand customers, having access to information concerning check numbers, sold items and paid amounts. The aim of the team was to understand the differences between tourist buyers and locals. They combined specific items for which tourists are potential buyers, such as Georgian wine, water and mineral water, napkins, nuts and dried fruits, Georgian cheese, vegetables, fruits, and national spices. According to their analysis, on average tourists purchase more items and therefore spend more in the same shop.

Inception

Taking into account that pricing policies are one of the most crucial factors for efficient businesses, this group analyzed the change in sales during a working day (Wednesday) and a weekend day (Sunday) in Spar supermarkets. The analysis was conducted for aggregated, store-specific and product group-specific data. The aim of the group was to analyze the sales tendency during these two different days. They showed that the distribution of the most demanded products does not differ much. The average sales in the shops on Wednesday reaches its maximal value during 8:00-9:00 pm, while the situation is smoother on Sunday.

The opening time and locations of SPAR branches is the second data set this group worked with. They created interactive maps for the branches (where and when the shops were opened), on which they demonstrated the time and locations of the openings of the branches using the skills they gained in Geospatial Data Analysis.

ISET is grateful for the opportunity its students had to expand their skillsets, and would like to thank the organizers, partners and mentors who helped to make the event possible.