newjibs · real estate investment simulator
Period
newjibs lets users invest a virtual budget in real estate and check returns two years later. It started from the steep rise in housing prices during COVID‑19. I built the backend, servers and CI/CD alone.
Sales records reduced from over 4 million to about 60,000
- Kakao Map cluster rendering by latitude and longitude
- Real estate information lookup
- Ranking
- Notices and real estate news
- JWT‑based member authentication
- Vue.js frontend deployed on Netlify
- Spring Boot, MariaDB and Redis containers shipped to Cloudtype and EC2 by GitHub Actions
Shrinking the sales data
- Problem
- Sale records since 2016 exceeded 4 million rows and slowed rendering, and simulated investing did not need all of them. Prices rose sharply after COVID‑19 began in 2020, so I used that as the basis for cutting the data.
- Solution
- Matched 2020 and 2022 sales one‑to‑one by floor and unit size, keeping price changes for the same homes. I judged that sales of the same home would make the investment results more trustworthy. Sales records fell from more than 4 million to about 60,000, and rendering became faster.
News crawling getting blocked
- Problem
- Collecting real estate news with jsoup on the hour got the server flagged as a bot and its IP blocked every time. I wanted to fix this inside the crawler rather than introduce Selenium.
- Solution
- Found that the on‑the‑hour timing and the request headers made the crawler look like a bot. The crawler now changes its request headers and adds a random delay before each call. With a less obvious call pattern, it stopped getting blocked.
Architecture
