General Purpose Machine Learning, Deep Learning And Reinforcement Learning For Roblox (Scikit-Learn-Like API + Business-Grade + For Revenue Generation)
A cloud-powered control dashboard that connects to Roblox games in real-time and enables you to build machine learning pipelines for proactive analytics inside your games.
Control your machine learning models inside Roblox’s servers through the front end
Take advantage of Roblox’s MemoryStoreService to reduce command fetching from the back end
Add custom commands and model names for your machine learning models
Visualize the models’ weights for data analysis
The Lua code for Roblox’s servers are 90% open source
Security
RSA-AES hybrid encryption between the front end and back end
Password hashing
HTTPS enforced
Use Cases
Proactive Analytics: Train ML models on Roblox servers and receive aggregated information on the dashboard.
Predictive Modelling: Predict trends from historical and real-time data
Live A/B Testing: Deploy variations of machine learning models on the go
Real-Time Data Monitoring: Visualize in-game data as it’s generated
Anomaly Detection: Detect outliers in player behaviour and game states
Note:
Currently, this system is a minimum viable product (MVP). That means a lot of features for paid plans are not added yet. Let me know if you are willing to pay for them.
Planned Pricing
Plan
Price/Month
API Keys
Commands Per API Key
Models Per API Key
Daily Command Requests
Features
Free
$0
1
3
3
10
Basic Machine Learning
Casual
$3
1
10
5
50
Basic Machine Learning, Scheduling
Indie
$15
3
20
10
250
Basic Machine Learning, Scheduling
Pro
$50
5
40
30
1,000
Basic Machine Learning, Scheduling, Model Tracking
Enterprise
$200
10
60
50
2,000
Basic Machine Learning, Scheduling, Model Tracking, Priority Support
Enterprise Plus
$500
20
100
100
5,000
Basic Machine Learning, Scheduling, Model Tracking, Priority Support
Which plans will you choose with your current budget?
Casual
Indie
Pro
Enterprise
Enterprise Plus
0voters
Which plans are you willing to pay if you have money for it? (Multiple Choice)
Roblox’s Open Cloud API is mainly concerned with Roblox’s game data and asset management that may or may not related to your games. As such, you might be able to integrate Roblox data and assets with discords bots or on your own website.
DataPredict Nucleus is a machine learning model management system, where you add and control machine learning models in your games. Once your game is live and is running DataPredict’s machine model, you can control these models using the DataPredict Nucleus website. These real-time control use cases includes:
Adjusting time-to-leave and probability-to-leave machine learning models.
Replacing non-functional machine learning models with others that are functional from other servers.
Real-time statistics on what is causing player to leave, or identifying the clusters that the players tend to aggregate on DataPredict Nucleus’ website.
Change machine learning models by manually setting it from the DataPredict Nucleus website.
As such, if you’re into proactive analytics and want to do machine learning optimize the game for you instead of manual intervention, you can use DataPredict Nucleus for that.
Currently, the database responsible for login is offline, but I can activate it at anytime. Let me know if you want to try it out.
The answer is quite dependent on what you’re trying to do. For starters, I recommend you to have a look at this DataPredict’s “High-Value Project Tutorials” here.
Since it is an MVP stage, it only support numerical values for now. However, should the demand somehow arise for text, image, audio or even model data, I will look into it.
Yes, the entire model can run inside Roblox servers. The performance impact largely depends on data size, and is multiplied by the factor of your model choice. However, small datasets are sufficient for training. DataPredict Nucleus is just doing the model management. DataPredict library itself (separate from DataPredict Nucleus) provides these models that you can run on Roblox’s servers.
None, since it is trained on Roblox’s server.
Again, dependent on the dataset size and model that you are using. Though, I could guarantee that one datapoint (not the whole dataset) takes 0.0001-0.001 seconds to train a model.
Yes. You can do this by having two models, where one is responsible for training and one is responsible for prediction. Every once in a while, you can transfer the model parameters to the prediction model without blocking the training model’s thread.
Edit: I misread the first question thanks to the fatigue. And to answer your question properly, we do not offer pretrained models for now. Though I am trying to create a community that has the culture of sharing pre-trained models.
Currently, the community is a bit inactive though since I made a mistake of not growing it by the time I first released DataPredict. I’ll probably focus on growing again since I am done with DataPredict development.
But if you’re not interested in inactive servers, we’re also cool with that too!
