DigiFi

digifi is a general purpose financial library and framework for financial modelling, portfolio optimisation, and asset pricing. The purpose of this library is to provide basic functions and algorithms used in finance, but with enhanced memory safety, minimum number of dependencies and low computational requirements. digifi is composed of the following modules:

utilities: Generic utility functions including time-value calculators, mathematical functions, loss functions and numerical solvers.
statistics: Contains probability distributions (e.g., Normal, Binomial, Poisson) along with methods for computing covariance, skewness, kurtosis, and gamma and beta functions.
random_generators: Algorithms for generating pseudo-random numbers from uniform and normal distributions, along with the algorithms that connect with probability distributions through a polymorphism to allow generation of pseudo-random numbers from any defined distribution.
stochastic_processes: Stochastic processes for simulation of price action, and a builder of custom stochastic processes.
financial_instruments: Pricing of different financial instruments (e.g., bonds, stocks, options) and their utility functions.
portfolio_applications: Portfolio performance metrics, risk metrics, algorithm for optimising portfolio of financial instruments for a given performance metric.
lattice_models: Binomial and trinomial models for pricing options (European, American and Bermudan) with any payoff (e.g., Call, Straddle).
monte_carlo: Function for pricing custom payoffs over any stochastic processes using Monte-Carlo simulations.
corporate_finance: Functions for valuation of investment projects, and CAPM.
technical_indicators: Trading indicators such as RSI, MACD, Bollinger Bands, etc.
market_making: Automated Market Making (AMM) algorithm for simulation of transactions in liquidity pools.

Disclaimer

Note that the developers of this package do not accept any responsibility or liability for the accuracy or completeness of the code or the information provided.

Features

The following optional features are provided by digifi (Note: none of these features are the enabled by default):

sample_data: Provides sample data to test CAPM and portfolio optimisation algorithms.
serde: Provides serialization/deserialization for certain structs.
plotly: Provides functions for plotting results using plotly library.

Errors

In general, the functions and methods inside digifi propagate errors so that the users can decide on the error handling techniques applicable to their use case. However, for some numerical solutions closures are used which can panic at runtime.

Citations

Some of the code and algorithms used are from external sources (aside from the dependencies of the library). Below is the list of the sources where the code extracts/algorithms were taken from, along with the purpose they serve in the library. We would like to express gratitude to the developers and academics who contributed to these projects.

Golden Ratio, Nelder-Mead and L-BFGS Numerical Solvers (https://github.com/to266/optimize): Unfortunately, the crate optimize is outdated, but it contains really good implementations of different numerical solvers. optimize is not a dependency to this library, but we have used extracts from its source code and slightly changed it to optimise it for our needs. We also reused some parts of their documentation as it was very well explained by its developers. Nonetheless, optimize is a great package and we hope there will be a renewed support for it.
1. Gradient Descent and Line Search Algorithms (https://github.com/aomader/optimization-rust): We have adapted the gradient descent and line search algorithms from the crate optimization for the same reasons as above. We appreciate the great code base of the package, but unfortunately it is not actively maintained, so we decided to adapt the code from it as aopposed to use it as our dependency.
Inverse CDF of Poisson Distribution (https://people.maths.ox.ac.uk/gilesm/codes/poissinv/paper.pdf): This paper covers the implementation of the method for approximating the inverse CDF of the Poisson distribution.
Algorithms for ln Gamma, ln Beta, Regularized Incomplete Beta and Inverse Regularized Incomplete Beta functions (https://github.com/statrs-dev/statrs): The implementation of these functions were taken from statrs.

General information

If you would like to add a commit or an issue, please do so using the GitHub link to the project:

https://github.com/Digital-Finance-DigiFi/digifi