decision-tool 1.0.0

Decision analysis package with solution and visualization.

Project description

Decision-tool is a simple yet effective decision analysis tool. Solves and visualizes Trees with recursive algorithm.

Usage

Decision trees are constructed using Tree class. it takes nested structure and returns a Tree instance. In the nested structure, dictinary represents decision, List represents discrete probability distribution and tuple represents possible outcome of a discrete probability distribution, scipy.stats can be used for continous probability distribution:

example_tree = Tree({"option 1":25,
                     "option 2":[(.4,20,"result 1"), (.3,21,"result 2"), (.3,stats.norm(1,1),"result 3")]
                     })
example_tree.fig().show()

Consider the example: You have 2 options as quit the job and not quit. if you quit, you will get a bonus of 8000$. If you do not quit, you will promote to manager with a probability of 0.3, If you then quit, will get 16000, If you do not, you will get a random payoff distributed normally with mean 12000 and standart deviation of 1000$. If you do not quit at the begining and do not promote, your PV will be normal random variable with mean 10000$ and standart deviation of 3000$.

job_decision = Tree({
    "quit":8e3,
    "do not quit":[(0.3, 
                        {"quit":16e3,
                        "do not quit":stats.norm(12e3,1e3)
                        }
                    ,"promote"),

                    (0.7, stats.norm(10e3,3e3),"do not promote")
                    ]
})

job_decision.fig().show()

The method solve returns a Tree instance that only contains optimal solutions.

solution = job_decision.solve()
print("optimal expected monetary value:" ,solution.mean,"variance:",solution.var)
solution.fig().show()

    

optimal expected monetary value: 11800.0 variance: 13860000.0

For indexing, order decision names and event names:

print("Variance of not quiting and not getting a promote ",
      solution["do not quit","do not promote"].var)

Variance of not quiting and not getting a promote  9000000.0

Save and load decision trees using save method and load_tree function.