# Enigma - Codefest’26 - IIT (BHU) Varanasi — Data

来源：https://www.kaggle.com/competitions/enigma26/data

采集时间：2026-10-19T21:23:14Z

# UPD: Submit your [code here](https://forms.gle/xuvwaz4CAG4hrfwq9) for prize money and participation certificates 

The dataset for this competition consists of three primary files: `train.xlsx`, `test.xlsx,` and `target.csv`. These files contain detailed information about participants in networking events and their algorithmically determined compatibility scores.

## File Descriptions

 - **`train.xlsx`**: Contains profiles of participants used for training your model. Each row represents a unique individual with various professional and personal attributes.
 - **`test.xlsx`**: Contains profiles of participants for whom you need to predict compatibility. The structure is identical to `train.xlsx`.
- **`target.csv`**: A pairwise compatibility file containing rows of the form (`src_user_id`, `dst_user_id`, `compatibility_score`), where each row denotes the compatibility score between two participants.

**Note on Compatibility:** These values were calculated algorithmically and do not necessarily represent "true" human compatibility. Participants are encouraged to develop a logical solution that goes beyond simple metrics like cosine similarity, incorporating the complex dynamics of the provided features.

`target.csv` consists of pairwise compatibility entries represented as triplets of the form (`src_user_id`, `dst_user_id`, `compatibility_score`). Each row denotes the algorithmically computed compatibility between a pair of participants. These values do not necessarily represent true real-world compatibility. Participants are encouraged to develop logical and robust solutions that go beyond simple similarity measures (such as cosine similarity) and incorporate richer interaction dynamics between participants.

## Columns in `train.xlsx` and `test.xlsx`

- **Profile_ID:** Id of participant
- **Age:** Age of participant
- **Gender:** Self explanatory
- **Role:** Position in his/her/they/them company.
- **Seniority_Level:** Level of the participant in his role
- **Company_Name:** Name of the company participant is working in.
- **Company_Size_Employees:** Number of employees in the company
- **Industry:** The domain of participant's profession.
- **Location_City:** The location of his company.
- **Business_Interests:** Multiple interests of the participant separated by semicolon (;)
- **Business_Objectives:** Multiple objectives of the participant separated by semicolon (;)
- **Constraints:** Multiple constraints placed by the participant separated by semicolon (;)

# Task objective: 

The objective of this competition is to predict compatibility scores between pairs of participants in professional networking events using the provided participant attributes. Given the detailed information available in `train.xlsx `and `test.xlsx`, participants are required to build models that can infer meaningful relationships between individuals and estimate their pairwise compatibility. The ground-truth compatibility values are provided in `target.csv` for training purposes.

A key challenge of this task is to move beyond naïve similarity-based approaches (such as direct cosine similarity on embeddings) and instead capture higher-order interactions, contextual alignment, and latent dynamics between participants. These may include, but are not limited to, shared professional goals, complementary roles, industry relevance, constraints, and organizational context.