Neo4J Graph Summit 2023



Last week I got the chance to attend Graph Summit 2023 Jakarta, organized by neo4j. This event discusses the power of the graph database compared to a traditional relational database in terms of complex relations and data queries.

There are also several presentations regarding the real-world use case from different speakers.

Influencer Taxes

One that is memorable for me is from Mr. Iwan Djuniardi about how the Directorate General of Taxes Indonesia tries to use Graph database to detect potential income of internet influencer, and thus calculating the tax.

The main problem is that one influencer might have different accounts across different platforms (i.e. Twitter, Facebook, Instagram, etc), or multiple accounts on one platform. Consolidating this information using a traditional approach might be too difficult and costly, but this type of relationship could be easily represented via a graph.

To go one higher level, they also use image segmentation to detect objects within a photo, to get a better understanding of their income level. Someone that posts a photo of an expensive Gucci Bag without tax information (i.e. never pay tax) should raise a red flag that will be checked by the teams at the Directorate General of Taxes.

Halal Food Probability

Another memorable presentation is from Mrs. Nur Aini Rakhmawati, an associate professor at Institut Teknologi Sepuluh Nopember Surabaya that use a graph database to predict the probability of a halal food based on its ingredients.

The way it works (based on my understanding) is by using graph in the preprocessing and feature selection process, to get the relationship between each ingredient. Since one ingredient could be composed of other ingredients, creating a graph relation, one could know using graph, whether all the ingredients that make it is from halal product or not. This information then could be fed up into another machine learning model as a prediction step.


Last but not least, I also had the chance to learn a little bit about the Neo4J query language called Chyper in the workshop. This is akin to SQL for the traditional database that we could use to get data out of Neo4J.

During the workshop, we are trying to predict fraud from some dummy transaction data. Albeit not fully complete from start to end, this workshop provide an introduction to some basic graph techniques that could be useful when dealing with graph data (so-called graph engineering).

Overall, this is a good event with so much insightful information for me. They have GraphMeetup event, which is a community-based event to share your project using Neo4j. Hopefully one day I could present something in this community.