To start the Bayesian network generator in Lisp:
1. Open your Lisp.
2. Load the file 'run.lisp'.
This will start the program, using configuration info from the file 'config.json'. A log
file will be written to the file 'log-chordalysis-TIMESTAMP.txt'. The resulting graph
will be written to 'graph-TIMESTAMP.py". A text file with graph data that can be loaded
into Mathematica will be written to 'graphical-log-TIMESTAMP.txt'.
Here's what the JSON configuration file can look like (except that you shouldn't '//' for
comments, because JSON doesn't have a syntax for comments).
-----------------------------------------------------------------------------------
{
// The name of the database or CSV file. Must be specified.
"datasource": "data/orphamine.csv"
// The maximum number of columns (or 'predicates', if you prefer).
// All other columns will be ignored.
"max_columns": 95,
// If a column has more than this number of distinct values, it is ignored.
"max_distinct_values": 50,
// Frequency tables with more than this number of entries are not computed and the
// corresponding edges will be skipped.
"max_frequency_table_size": 10000000,
// The system will keep adding edges to the graph until either [1] 'max_iterations' is
// reached, or [2] there are no edges left with a p-value of at least 'threshold'.
"max_iterations": 30,
"threshold": 0.05,
// A list with values that should be considered to be missing values.
"default_missing": ["", "-"],
// A list of column specifications. Each specification can include the following parts:
//
// "name" [required]: the field name
//
// "missing" [optional]: a list of values that should be interpreted as missing value
// (in addition to the ones specified by 'default_missing'). Example: ["999"].
//
// "type" [optional, default: "categorical"]: either "categorical" or "numerical".
//
// "quantiles" [optional]: may be specified if the field type is "numerical".
// If specified, this is a list of numbers that divide the quantile intervals.
// Example: {0,10,20,50,100} will create 4 quantiles: 0 <= x < 10, 10 <= x < 20,
// 20 <= x < 50, 50 <= x < 100.
"columns":
[{"name": "tot_money_ind",
"type": "numerical",
"quantiles": [0, 500, 1000, 2000, 3000, 5000, 10000]},
{"name": "max_money_ind",
"type": "numerical"},
{"name": "age",
"type": "numerical",
"quantiles": [10, 20, 30, 40, 60, 80]},
{"name": "tot_score_ind",
"type": "numerical",
"missing": ["", "0.0"],
"quantiles": [1, 10, 20, 30, 50, 100]}
]
}
----------------------------------------------------------------------------------------