using DataFrames, RDatasets

vn = dataset("Ecdat","VietNamI")

	Pharvis	LnhhExp	Age	Sex	Married	Educ	Illness	Injury	Illdays	Actdays	Insurance	Commune
1	0	2.730363	3.7612	male	1	2	1	0	7	0	0	192
2	0	2.737248	2.944439	female	0	0	1	0	4	0	0	167
3	0	2.266935	2.56495	male	0	4	0	0	0	0	1	76
4	1	2.392753	3.637586	female	1	3	1	0	3	0	1	123
5	1	3.105335	3.295837	male	1	3	1	0	10	0	0	148
6	0	3.760884	3.367296	male	1	9	0	0	0	0	1	20
7	0	3.155609	3.663562	female	1	2	0	0	0	0	1	40
8	0	3.724682	2.197225	male	0	5	0	0	0	0	1	57
9	2	2.861691	3.7612	female	1	2	2	0	4	0	0	49
10	3	2.615077	4.234107	male	1	0	1	0	7	0	0	170
11	1	2.653243	2.772589	male	0	4	1	0	1	0	0	40
12	1	2.139857	3.663562	female	1	2	2	0	5	0	0	127
13	2	2.625683	3.555348	female	1	3	1	0	3	0	0	106
14	1	2.767746	1.94591	female	0	5	2	0	3	0	0	168
15	0	2.871242	2.302585	male	0	0	0	0	0	0	0	185
16	2	1.983811	3.135494	male	1	4	3	0	10	0	0	41
17	0	1.260201	3.091043	male	0	4	0	0	0	0	0	106
18	0	2.298178	3.401197	male	1	2	0	0	0	0	0	141
19	3	2.132508	3.332205	male	1	4	3	0	10	0	0	61
20	0	2.000231	3.258096	male	1	3	0	0	0	0	0	159
21	0	2.018375	3.218876	male	1	6	1	0	7	0	0	121
22	1	1.886142	2.639057	female	0	2	1	0	3	0	0	56
23	10	2.953125	1.098612	female	0	2	2	0	4	0	0	34
24	0	2.378481	3.526361	male	1	0	0	0	0	0	0	127
25	0	1.574376	3.555348	male	1	2	0	0	0	0	0	135
26	3	2.495514	3.637586	female	1	3	1	0	5	0	0	147
27	1	2.318077	3.7612	male	1	4	1	0	5	0	0	94
28	0	2.029045	4.248495	female	1	3	1	0	30	0	0	125
29	1	1.788754	3.610918	female	1	3	3	0	3	0	0	79
30	0	2.091107	2.079442	female	0	0	0	0	0	0	0	143
⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮

Predicting "Days Ill" based on historic data

days_ill = vn[:Illdays]

27765-element DataArray{Int32,1}:
  7
  4
  0
  3
 10
  0
  0
  0
  4
  7
  1
  5
  3
  ⋮
  3
 10
  0
  1
  0
  0
 30
  3
  0
  0
 20
  7

We've isolated the signal we want to analyze, now we remove it from the feature set

delete!(vn,:Illdays)

	Pharvis	LnhhExp	Age	Sex	Married	Educ	Illness	Injury	Actdays	Insurance	Commune
1	0	2.730363	3.7612	male	1	2	1	0	0	0	192
2	0	2.737248	2.944439	female	0	0	1	0	0	0	167
3	0	2.266935	2.56495	male	0	4	0	0	0	1	76
4	1	2.392753	3.637586	female	1	3	1	0	0	1	123
5	1	3.105335	3.295837	male	1	3	1	0	0	0	148
6	0	3.760884	3.367296	male	1	9	0	0	0	1	20
7	0	3.155609	3.663562	female	1	2	0	0	0	1	40
8	0	3.724682	2.197225	male	0	5	0	0	0	1	57
9	2	2.861691	3.7612	female	1	2	2	0	0	0	49
10	3	2.615077	4.234107	male	1	0	1	0	0	0	170
11	1	2.653243	2.772589	male	0	4	1	0	0	0	40
12	1	2.139857	3.663562	female	1	2	2	0	0	0	127
13	2	2.625683	3.555348	female	1	3	1	0	0	0	106
14	1	2.767746	1.94591	female	0	5	2	0	0	0	168
15	0	2.871242	2.302585	male	0	0	0	0	0	0	185
16	2	1.983811	3.135494	male	1	4	3	0	0	0	41
17	0	1.260201	3.091043	male	0	4	0	0	0	0	106
18	0	2.298178	3.401197	male	1	2	0	0	0	0	141
19	3	2.132508	3.332205	male	1	4	3	0	0	0	61
20	0	2.000231	3.258096	male	1	3	0	0	0	0	159
21	0	2.018375	3.218876	male	1	6	1	0	0	0	121
22	1	1.886142	2.639057	female	0	2	1	0	0	0	56
23	10	2.953125	1.098612	female	0	2	2	0	0	0	34
24	0	2.378481	3.526361	male	1	0	0	0	0	0	127
25	0	1.574376	3.555348	male	1	2	0	0	0	0	135
26	3	2.495514	3.637586	female	1	3	1	0	0	0	147
27	1	2.318077	3.7612	male	1	4	1	0	0	0	94
28	0	2.029045	4.248495	female	1	3	1	0	0	0	125
29	1	1.788754	3.610918	female	1	3	3	0	0	0	79
30	0	2.091107	2.079442	female	0	0	0	0	0	0	143
⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮

# Pkg.add("DecisionTree")
using DecisionTree

The DecisionTree package works with Julia Arrays, so we make those conversions

signals = convert(Array,days_ill)
features = convert(Array,vn);

methods(build_forest)

5 methods for generic function build_forest:

• build_forest{T<:FloatingPoint,U<:Real}(labels::Array{T<:FloatingPoint,1},features::Array{U<:Real,2},nsubfeatures::Integer,ntrees::Integer) at /home/juser/.julia/v0.3/DecisionTree/src/DecisionTree.jl:415
• build_forest{T<:FloatingPoint,U<:Real}(labels::Array{T<:FloatingPoint,1},features::Array{U<:Real,2},nsubfeatures::Integer,ntrees::Integer,maxlabels) at /home/juser/.julia/v0.3/DecisionTree/src/DecisionTree.jl:415
• build_forest{T<:FloatingPoint,U<:Real}(labels::Array{T<:FloatingPoint,1},features::Array{U<:Real,2},nsubfeatures::Integer,ntrees::Integer,maxlabels,partialsampling) at /home/juser/.julia/v0.3/DecisionTree/src/DecisionTree.jl:415
• build_forest(labels::Array{T,1},features::Array{T,2},nsubfeatures::Integer,ntrees::Integer) at /home/juser/.julia/v0.3/DecisionTree/src/DecisionTree.jl:245
• build_forest(labels::Array{T,1},features::Array{T,2},nsubfeatures::Integer,ntrees::Integer,partialsampling) at /home/juser/.julia/v0.3/DecisionTree/src/DecisionTree.jl:245

Random Forest

build_forest parameters are:

signals _array of the signal we want to calculate
features the corresponding feature array that indicates those signals
festures used the number features for the each split or branch of the tree
number of trees trees in the forrest, larger takes longer, but could be more accurate.
sampling rate number lowered from 1.0 to favor minority signals

model = build_forest(signals,features,int(sqrt(length(features[1,:]))),10,.9)

Ensemble of Decision Trees
Trees:      10
Avg Leaves: 3674.7
Avg Depth:  24.3

Apply the algorithm to any 2D Matix of features

predictions = apply_forest(model,features)

27765-element Array{Any,1}:
  7
  3
  0
  3
 10
  0
  0
  0
  4
  7
  2
  5
  3
  ⋮
  3
 10
  0
  1
  0
  0
 30
  3
  0
  0
 20
  7

confusion_matrix(signals,predictions)

32x32 Array{Int64,2}:
 16566    0     5    10     1    1    0  …  0  0  0   0  0  0    0  0    1  0
     1  290    68    68    16    5    0     0  0  0   0  0  0    0  0    0  0
     1    4  1383   114    41    7    0     0  0  0   1  0  0    0  0    0  0
     0    2    87  1856    45    7    2     0  0  0   0  0  0    1  0    0  0
     1    1    32    95  1129    4    1     0  0  0   1  0  0    1  0    0  0
     1    0    56   143    46  841    1  …  0  0  0   0  0  0    1  0    0  0
     1    0    23    55    26    8  287     0  0  0   0  0  0    2  0    0  0
     1    0    24    66    16    8    0     0  0  0   0  0  0    0  0    0  0
     0    0     4    16    12    3    1     0  0  0   0  0  0    0  0    0  0
     0    0     1     6     0    0    0     0  0  0   0  0  0    0  0    0  0
     0    0    12    40    18    5    1  …  0  0  0   0  0  0    0  0    0  0
     0    0     0     1     0    1    0     0  0  0   0  0  0    0  0    0  0
     1    0     2     4     6    1    0     0  0  0   0  0  0    1  0    0  0
     ⋮                           ⋮       ⋱            ⋮                  ⋮   
     1    3     7    23    13    6    0  …  0  0  0   1  0  0    3  0    0  0
     0    0     0     1     1    0    0     0  0  0   0  0  0    0  0    0  0
     0    0     0     0     0    0    0     1  0  0   0  0  0    0  0    0  0
     0    0     0     0     0    0    0     0  3  0   0  0  0    0  0    0  0
     0    0     0     0     0    0    0     0  0  3   0  0  0    0  0    0  0
     0    0     0     0     1    2    0  …  0  0  0  56  0  0    0  0    0  0
     0    0     1     1     0    1    0     0  0  0   0  2  0    0  0    0  0
     0    0     0     0     0    0    0     0  0  0   0  0  3    0  0    0  0
     0    0    13    22     4    3    0     0  0  0   0  0  0  274  0    0  0
     0    0     0     0     0    0    0     0  0  0   0  0  0    0  1    0  0
     0    1     7    18     6    4    1  …  0  0  0   0  0  0    0  0  153  0
     0    0     0     0     1    0    0     0  0  0   0  0  0    0  0    0  0

Classes:  {0,1,2,3,4,5,6,7,8,9  …  22,23,24,25,26,27,28,29,30,60}
Matrix:   
Accuracy: 0.9280388978930308
Kappa:    0.8846820611436211

https://github.com/bensadeghi/DecisionTree.jl/tree/83765089feb5b2d30d72046ab78ca044841c827d
http://bensadeghi.com/decision-trees-julia/
http://appliedpredictivemodeling.com/blog/2013/12/8/28rmc2lv96h8fw8700zm4nl50busep