andrealmeida/erlang-decision-tree
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Compare commits

base: andrealmeida:master
Branches Tags
andrealmeida:master
andrealmeida:PATCH
andrealmeida:MINOR
andrealmeida:MAJOR
..
compare: andrealmeida:MINOR
Branches Tags
andrealmeida:master
andrealmeida:PATCH
andrealmeida:MINOR
andrealmeida:MAJOR

No commits in common. "master" and "MINOR" have entirely different histories.
master ... MINOR

4 changed files with 28 additions and 163 deletions
Show Stats Expand all files Collapse all files

91
README.md
View File

@ -1,64 +1,28 @@
# Erlang Decision Tree and Baysian Networks # Erlang Decision Tree and Baysian Networks
![alt text](https://img.shields.io/badge/version-0.0.1-success.svg)
This repository have two different exercises in erlang: This repository have two different exercises in erlang:
**1.Decision tree that maximizes the optimal point and consequently provides decision support based on the given assumptions.** **1.Decision tree that maximizes the optimal point and consequently provides decision support based on the assumptions provided.**
* The probability of taking medication/treatment 1 and the user being OK is 42%; * The probability of making a medication/treatment 1 and use is OK of 42%;
* The probability of taking medication/treatment 1 and the user becoming KO is 22.5%; * One possibility of making a medication/treatment 1 and of the user staying KO is 22.5%;
* The probability of taking medication/treatment 1 and it not being conclusive is 35.6%; * The probability of making a medication/treatment 1 and is not conclusive is 35.6%;
* The probability of taking medication/treatment 2 and being OK is 62%; * The probability of making a medication/treatment 2 and staying OK is 62%;
* The probability of taking a medication/treatment 2 and getting KO is 38%. * The probability of making a medication/treatment 2 and getting KO is 38%.
``` ![alt text](https://i.postimg.cc/5207cpvs/Captura-de-ecr-2019-04-09-s-13-06-01.png)
-module(engine).
-compile([export_all]).
% Decision Tree
%
% +----+
% +---> OK |
% | +----+
% |
% +----------| +----+
% | DO MED 1 +---> KO |
% +----------+ +----+ +----+
% | +---> OK |
% | +---------------| +----+
% +---> DO MED 2 +
% +---------------| +----+
% +---> KO |
% +----+
% Each edge can be calculated from the table
doAll() ->
T = dec_tree:start(),
T ! {create_node, 'DO MED1', 0, none},
T ! {create_node, 'MED1 OK', 0.980, 'DO MED1'},
T ! {create_node, 'MED1 KO', 0.225, 'DO MED1'},
T ! {create_node, 'DO MED2', 0.356, 'DO MED1'},
T ! {create_node, 'MED 2 OK', 0.62, 'DO MED2'},
T ! {create_node, 'MED 2 KO', 0.38, 'DO MED2'},
T ! {solve}.
```
**2. There are ways of obtaining probabilities knowing certain parameters (they don't indicate the optimum point, but calculate a probability of success depending on the existing resources) - using Baysian networks.** **2.There are ways to get the probabilities knowing certain parameters (they do not indicate the optimal point, but calculate a probability of success depending on the existing resources) - using the baysian networks.**
What is the probability of giving the patient M2 and the patient being well, knowing that: What is the probability of giving the patient M2 and is OK, knowing that:
* The probability of administering M1 and the patient being well is 20 per cent; * The probability of administering M1 and the patient is OK is 20%;
* The probability of administering M2 knowing that I have administered M1 and the patient is well is 70%; * The probability of administering M2 knowing that I administered M1 and the patient being OK is 70%;
* The probability of administering M2 knowing that I have not administered M1 and the patient is OK is 20 per cent. * The probability of administering M2 knowing that I did not administer M1 and the patient is OK of 20%.
### Prerequisites ### Prerequisites
@ -69,12 +33,12 @@ Basic knowledge of erlang, algorithm and statistics.
### Install or use Docker ### Install or use Docker
If you prefer to install the erlang compiler, please find the appropriate installation for your operating system. If you prefer install the erlang compiler, please search for the appropriate installation for your OS.
If you prefer, you can use Docker to test this solution. If you prefer you can use Docker to test this solution.
Get a [docker container](https://hub.docker.com/search?q=erlang&type=image) and start the container. Get a [docker container](https://hub.docker.com/search?q=erlang&type=image) and start the container:
I suggest [this container](https://hub.docker.com/r/bitwalker/alpine-erlang) and run:
I suggest [this container](https://hub.docker.com/r/bitwalker/alpine-erlang) and run it:
``` ```
docker pull bitwalker/alpine-erlang docker pull bitwalker/alpine-erlang
@ -86,15 +50,15 @@ docker run --rm -it --user=root bitwalker/alpine-erlang
1. For the first exercise (Decision Tree): 1. For the first exercise (Decision Tree):
``` ```
c(motor). c(engine).
motor:doAll(). engine:doAll().
``` ```
and the best solution is: and the best solution is:
![alt text](https://i.postimg.cc/mg5dTMyB/Captura-de-ecr-2019-04-09-s-13-05-46.png) ![alt text](https://i.postimg.cc/mg5dTMyB/Captura-de-ecr-2019-04-09-s-13-05-46.png)
** You can take medication 1 and medication 2 with a 97.6 per cent success rate to get well. ** You can do a medication 1 and medication 2 of 97.6% of success to be OK.**
@ -103,16 +67,16 @@ and the best solution is:
2. For the second exercise (Baysian Networks): 2. For the second exercise (Baysian Networks):
``` ```
c(motor3). c(engine3).
motor3:getp({m2, ok}). engine3:getp({m2, ok}).
``` ```
![alt text](https://i.postimg.cc/QxBm0Vrz/Captura-de-ecr-2019-04-09-s-15-57-38.png) ![alt text](https://i.postimg.cc/QxBm0Vrz/Captura-de-ecr-2019-04-09-s-15-57-38.png)
**There is a 16 per cent chance that the patient will be OK when administering drug M2. **There is a 16% chance that the patient will be OK administering the drug M2.**
## Created with ## Built With
* erlang - https://www.erlang.org; * erlang - https://www.erlang.org;
@ -122,12 +86,13 @@ Please read [CONTRIBUTING.md](CONTRIBUTING.md) for details on my code of conduct
## Versioning ## Versioning
I use [SemVer](http://semver.org/) for versioning. For available versions, see [tags on this repository](https://gitlab.andrealmeida.net/public_projects/erlang-decision-tree/tags). I use [SemVer](http://semver.org/) for versioning. For the versions available, see the [tags on this repository](https://gitlab.andrealmeida.net/public_projects/erlang-decision-tree/tags).
## Authors ## Authors
[![Author](https://img.shields.io/badge/Author-André%20Almeida-blue.svg)](https://andrealmeida.net) * **André Almeida** - [andrealmeida.net](https://andrealmeida.net)
## Licence ## License
This project is licensed under the MIT License - see the [LICENSE.md](LICENSE.md) file for details
This project is licensed under the MIT licence - see the file [LICENSE.md](LICENSE.md) for more details.

29
app/baysian-network/engine3.erl
View File

@ -1,29 +0,0 @@
-module(engine3).
-compile([export_all]).
events() -> [{m1, ok}, {m2, ok}].
p({m1, ok}) -> 0.2;
p(A) when is_tuple(A) -> none.
cp({m2, ok}, {m1, ok}) -> 0.8;
cp(_, _) -> none.
getp({A, B}) when is_tuple(A), is_tuple(B) ->
case cp(A, B) of
none ->
Pba = cp(B, A),
Pa = getp(A),
Pb = getp(B),
Pba * Pa / Pb;
X ->
X
end;
getp(A) when is_tuple(A) ->
case p(A) of
none ->
lists:sum([ getp({A, B})*getp(B) || B <- events(), p(B) /= none, cp(A,B) /= none ]);
X ->
X
end.

41
app/decision-tree/dec_tree.erl
View File

@ -1,41 +0,0 @@
-module(dec_tree).
-export([start/0, decision_tree/0]).
%%% Interface
start() -> Pid = spawn(?MODULE, decision_tree, []), Pid.
decision_tree() -> T = create_tree(), decision_tree(T).
decision_tree(Tree) ->
io:format("~w~n", [Tree]), %% for debug purposes we always print the Tree
receive
{create_node, Node, Edge, Parent} -> NewTree = create_node(Tree, Node, Edge, Parent), decision_tree(NewTree);
{print} -> decision_tree(Tree); % no need to print the tree, since we already printed before receive...
{solve} -> io:format("Solution: ~w~n", [solve(Tree)]), decision_tree(Tree)
end.
%%% Create the decision Tree
create_tree() -> none.
create_node(none, Node, Edge, _) -> {Node, Edge, []}; % 1st node
create_node({Parent, P_Edge, L}, Node, Edge, Parent) -> {Parent, P_Edge, [{Node, Edge, []}|L]}; % we find the parent node and insert the new node
create_node({Root, R_Edge, []}, _, _, Parent) -> {Root, R_Edge, []}; % recursion stop when the parent node is not found
create_node({Root, R_Edge, L}, Node, Edge, Parent) -> {Root, R_Edge, lists:map(fun(N) -> create_node(N, Node, Edge, Parent) end, L)}. % we try to insert the node in all subtrees
%%% Solve the decision Tree
% Determines the path by choosing the greatest edge
solve_max([{Edge, Path}]) -> {Edge, Path};
solve_max([{H_Edge, E_Path}|T]) ->
{T_max, T_Path} = solve_max(T),
if
H_Edge > T_max ->
{H_Edge, E_Path};
true -> % works as an 'else' branch
{T_max, T_Path}
end.
solve(none) -> false; %there is no solition for an empty decision tree
solve({Node, Edge, []}) -> {Edge,[Node]}; %recursion stop
solve({Node, Edge, L}) -> {C_Edge, C_Path} = solve_max(lists:map(fun(N) -> solve(N) end, L)), {Edge + C_Edge, [Node|C_Path]}.

30
app/decision-tree/engine.erl
View File

@ -1,30 +0,0 @@
-module(engine).
-compile([export_all]).
% Decision Tree
%
% +----+
% +---> OK |
% | +----+
% |
% +----------| +----+
% | DO MED 1 +---> KO |
% +----------+ +----+ +----+
% | +---> OK |
% | +---------------| +----+
% +---> DO MED 2 +
% +---------------| +----+
% +---> KO |
% +----+
% Each edge can be calculated from the table
doAll() ->
T = dec_tree:start(),
T ! {create_node, 'DO MED1', 0, none},
T ! {create_node, 'MED1 OK', 0.980, 'DO MED1'},
T ! {create_node, 'MED1 KO', 0.225, 'DO MED1'},
T ! {create_node, 'DO MED2', 0.356, 'DO MED1'},
T ! {create_node, 'MED 2 OK', 0.62, 'DO MED2'},
T ! {create_node, 'MED 2 KO', 0.38, 'DO MED2'},
T ! {solve}.