Stock Optimize

Introduction

Stock Optimize is a Python project that implements various optimization algorithms to solve the portfolio optimization problem. The goal is to determine the optimal allocation of assets in a stock portfolio to maximize expected returns while minimizing risk.

This project was completed as part of the Artificial Intelligence course taken in the Fall of 2024 at Ferdowsi University of Mashhad.

Algorithms Implemented

The project includes the following optimization algorithms:

• Genetic Algorithm
• Beam Search
• Random Beam Search
• Simulated Annealing

These algorithms are defined in the Optimizer module and can be utilized to find the optimal investment strategy.

Installation

Ensure you have Python 3 installed. Install the required packages using:

pip install numpy matplotlib

Project Background

This project was completed as part of the Artificial Intelligence course in the Fall of 2024. It applies various optimization algorithms to solve the portfolio optimization problem, focusing on maximizing expected returns while minimizing risk.

Classes and Modules Description

Optimizer Module

The Optimizer module contains implementations of different optimization algorithms.

Vector Class

Represents an investment vector in the optimization process.

• Attributes:

  • __covariance_matrix: Covariance matrix of asset returns.
  • __expected_return: Expected returns for each asset.
  • __max_comparison: Boolean indicating comparison mode for fitness.
  • __number_children: Number of children vectors in certain algorithms.
  • __values: Allocation proportions for each asset.
  • __fitness: Fitness value of the vector.

• Methods:

  • __init__(self, vector): Initializes the vector with allocation values and calculates fitness.
  • __eq__(self, other): Checks equality between two vectors.
  • __lt__(self, other): Defines less-than comparison based on fitness.
  • values(self): Getter and setter for __values, ensures allocations sum to 1.
  • fitness(self): Returns the fitness value.
  • len(cls): Class method to get the number of assets.
  • max_comparison(cls, boolean_value): Sets comparison mode for fitness evaluation.
  • set_class_variables(cls, expected_return, covariance_matrix): Sets class-level variables.
  • __fitness_func(self): Calculates the fitness as return-to-risk ratio.
  • get_neighbours(self, learning_rate): Generates neighbouring vectors for exploration.

Optimizer Base Class

An abstract base class defining the structure for optimization algorithms.

• Attributes:

  • _num_iterations: Number of iterations for the optimizer.

• Methods:

  • optimize(self): Abstract method to perform optimization.
  • get_plot_info(self): Abstract method to retrieve data for plotting.

Optimization Algorithms

GeneticAlgorithmOptimizer Class

Implements the Genetic Algorithm for optimization.

• Attributes:

  • __num_individuals: Population size.
  • __population: List of current vectors in the population.
  • __iterations: Fitness values over iterations.
  • __mutation_normal_deviation: Standard deviation for mutation scaling.

• Methods:

  • __init__(self, num_individuals, num_iterations, random_state): Initializes the optimizer.
  • optimize(self): Executes the genetic algorithm steps: selection, crossover, and mutation.
  • get_plot_info(self): Provides data for plotting convergence.

BeamSearchOptimizer Class

Implements Beam Search optimization.

• Attributes:

  • __beam_length: Number of top vectors to retain each iteration.
  • __beam: Current list of vectors in the beam.
  • __LEARNING_RATE: Step size for generating neighbours.
  • __iterations: Fitness values over iterations.

• Methods:

  • __init__(self, beam_length, num_iterations, learning_rate, random_state): Initializes the optimizer.
  • optimize(self): Iteratively expands and selects the best neighbours.
  • get_plot_info(self): Provides data for plotting convergence.

RandomBeamSearchOptimizer Class

Implements Random Beam Search optimization, a variation of Beam Search.

• Attributes and Methods:

  Similar to BeamSearchOptimizer, but selection of beams is based on probabilistic sampling rather than deterministic ranking.

SimulatedAnnealingOptimizer Class

Implements Simulated Annealing optimization.

• Attributes:

  • __current_state: Current solution vector.
  • __LEARNING_RATE: Step size for neighbour generation.
  • __neighbours: Tuple containing the current state and its neighbours.
  • __iterations: Fitness values over iterations.

• Methods:

  • __init__(self, num_iterations, learning_rate, random_state): Initializes the optimizer.
  • optimize(self): Performs optimization using a temperature schedule to accept or reject new states.
  • __perturbate(self): Generates a new candidate solution.
  • __accept_change(self, new_state, temperature): Determines acceptance of new state based on probability.
  • get_plot_info(self): Provides data for plotting convergence.

Main Module

The main.py script coordinates the execution of the optimization algorithms.

• Functions:

  • combine_plots(optimizers): Plots the convergence of each optimizer for comparison.

• Workflow:

  • Reads input data from input.txt (expected returns and covariance matrix).
  • Sets class variables for the Vector class.
  • Initializes each optimizer with specified parameters.
  • Runs the optimization algorithms and outputs the optimal allocations.
  • Visualizes the performance of each algorithm using plots.