Tempus

Tempus is a project I've been working on for the past 10 years with the support of about 20 experts, professors and engineers from Belarus, Bulgaria, Czech Republic, Israel, Northern Macedonia, Russia and Ukraine focused on time-series analysis in near real-time. It aims to produce the highest precision forecast of time-series data on the market. While the source code is open and licensed under the BSD4 license, I offer paid consulting services for real-world application and a team of engineers prepared for maintenance and feature implementation according to client's needs.

So far we have accomplished the following:

• With the help of prof. Atanasov and ChatGPT we've corrected the support vectors machine theory by professors Vapnik and Chervonenkis allowing it to scale by nesting kernel matrices. Even with a single level of depth our SVM implementation has better precision for regression forecasts compared to a Microsoft LightGBM model - one of the best models for regression problems available today.

• This novel implementation of a support vectors machine can use any model for its kernel method, eg. a temporal fusion transformer, gradient boosted decision trees or a conventional kernel distance function like the RBF, global alignment or the path kernel, or even use another SVM as a kernel function. Scale it as much as you are willing to commit hardware resources to modeling or until you are satisfied with the forecast precision and all relevant information is extracted from the available data. Good fitness, further referred to as accuracy for readers convenience, significantly better than any modeling technique I'm aware of, can be established even from miniscule amounts of data presented to the software. This is accomplished by calculating the ideal kernel matrix for a given set of labels and training another model on the ideal kernel matrix itself as a dataset - the manifold.

• Beside nesting kernels, Tempus can scale in the time domain (using slicing) or spectral domain (using STFT, VMD or EMD decomposition), depending on how much hardware resources you are willing to commit to modelling your data. Sequential (residuals) boosting is also partially implemented (WIP). To counter extremely noisy data this SVM implementation supports multiple layers of weights used by the internal matrix solver. For real-time and near real-time purposes, online learning and forgetting is not fully implemented but it's in the works.

• Data connectors to FIX using QuickFIX (WIP), PostgreSQL, DDB and MQL5 for financial application or general purpose data source are available.

• The project is mainly implemented in C++, CUDA, OpenCL, OpenMP and MPI used for multiple computing nodes. Tempus can scale to many GPUs and many times more CPU cores per every computing node. Tempus only supports running on the Linux operating system.

• The project costed about 1.2 million euros on salaries, running expenses and development hardware for Tempus over the past 10 years. 

Investment in Tempus would allow me to purchase a higher quality solver (eg. Baron or KNitro), build more data connectors or permit hiring developers and domain experts (eg. traders) in order to apply Tempus to financial market forecasts, or any other application that may seem suitable to you. These are options to improve the quality of your custom Tempus application we tailor to your needs.

The Github repository is here, you can look at the sources and documentation, or request and I will permit you commit to the project. 

The former investor's servers were confiscated by Europol as the company was implied in financial malversations so a big part of the documentation was lost (tickets, manuals etc) but the source code is all here and I'm rewriting the documentation on the go. I managed to salvage everything critical so I can say it's operational and usable at this point.

I'm the lead developer as well as its founder and am continuously actively contributing to it. I provide consultancy services in setting up Tempus and exploiting it for your purposes. If anyone is interested in getting involved in the project or using it for their own purposes feel free to contact me. Also, you can see more details about Tempus and machine learning related topics on the statistical learning page.