Development of AI Model to Detect Cheaters in Rust Based on Player Statistics -- 2

I need to develop and train a neural network model to detect cheaters in the online game Rust based on player statistics. We have JSON files containing player statistics collected from several game servers. The statistics reset twice a week (wipes), and historical data is saved in JSON files. Each server logs its player statistics separately.

Data Details:
JSON File Structure: Each file contains a list of players with the following data points:
- `PlayerID`
- `PlayerName`
- `ShotsFired`
- `ShotsHit`
- `TotalHitDistance`
- `HitsHead`
- `HitsBody`
- `HitsLegs`
- `AimedShots`
- `AimedHits`
- `Kills`
- `Deaths`
- `Suicides`

Additional Data:
- List of all banned cheaters players with ban dates.
- List of all manually verified players with verification dates.
- Date of each statistics file.
- Approximately 5000 records per wipe period.
- Around 10-30 bans and verifications per wipe period.

Requirements:
- Use neural networks for classification.
- The model should be trained on provided data and able to detect cheaters players.
- Model must be accurate and efficient, working in production on CPU.
- The model should be designed to be updatable with new data as it becomes available.

Future Improvements:
- Provide recommendations on additional data that can be collected to improve model accuracy.
- Suggestions for continuous model improvement based on new data.