forked from zero-one-group/geni
/
regression.clj
162 lines (152 loc) · 7.11 KB
/
regression.clj
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
(ns zero-one.geni.ml.regression
(:require
[potemkin :refer [import-fn]]
[zero-one.geni.docs :as docs]
[zero-one.geni.interop :as interop]
[zero-one.geni.utils :refer [coalesce]])
(:import
(org.apache.spark.ml.regression AFTSurvivalRegression
DecisionTreeRegressor
FMRegressor
GBTRegressor
GeneralizedLinearRegression
IsotonicRegression
LinearRegression
RandomForestRegressor)))
(defn linear-regression [params]
(let [defaults {:max-iter 100,
:tol 1.0E-6,
:elastic-net-param 0.0,
:reg-param 0.0,
:aggregation-depth 2,
:fit-intercept true,
:label-col "label",
:standardization true,
:epsilon 1.35,
:loss "squaredError",
:prediction-col "prediction",
:features-col "features",
:solver "auto"}
std (coalesce (:standardisation params)
(:standardization params)
(:standardization defaults))
props (-> defaults
(merge params)
(assoc :standardization std))]
(interop/instantiate LinearRegression props)))
(defn generalized-linear-regression [params]
(let [defaults {:max-iter 25,
:variance-power 0.0,
:family "gaussian",
:tol 1.0E-6,
:reg-param 0.0,
:fit-intercept true,
:label-col "label",
:prediction-col "prediction",
:features-col "features",
:solver "irls"}
props (merge defaults params)]
(interop/instantiate GeneralizedLinearRegression props)))
(defn decision-tree-regressor [params]
(let [defaults {:max-bins 32,
:min-info-gain 0.0,
:impurity "variance",
:cache-node-ids false,
:seed 926680331,
:label-col "label",
:leaf-col ""
:checkpoint-interval 10,
:min-weight-fraction-per-node 0.0,
:max-depth 5,
:max-memory-in-mb 256,
:prediction-col "prediction",
:features-col "features",
:min-instances-per-node 1}
props (merge defaults params)]
(interop/instantiate DecisionTreeRegressor props)))
(defn random-forest-regressor [params]
(let [defaults {:max-bins 32,
:subsampling-rate 1.0,
:min-info-gain 0.0,
:impurity "variance",
:min-weight-fraction-per-node 0.0,
:cache-node-ids false,
:seed 235498149,
:label-col "label",
:leaf-col ""
:feature-subset-strategy "auto",
:checkpoint-interval 10,
:max-depth 5,
:max-memory-in-mb 256,
:prediction-col "prediction",
:features-col "features",
:min-instances-per-node 1,
:num-trees 20}
props (merge defaults params)]
(interop/instantiate RandomForestRegressor props)))
(defn gbt-regressor [params]
(let [defaults {:max-bins 32,
:subsampling-rate 1.0,
:max-iter 20,
:step-size 0.1,
:min-info-gain 0.0,
:cache-node-ids false,
:seed -131597770,
:label-col "label",
:leaf-col ""
:min-weight-fraction-per-node 0.0,
:feature-subset-strategy "all",
:checkpoint-interval 10,
:loss-type "squared",
:max-depth 5,
:max-memory-in-mb 256,
:prediction-col "prediction",
:features-col "features",
:min-instances-per-node 1}
props (merge defaults params)]
(interop/instantiate GBTRegressor props)))
(defn aft-survival-regression [params]
(let [q-probs [0.01, 0.05, 0.1, 0.25, 0.5, 0.75, 0.9, 0.95, 0.99]
defaults {:max-iter 100,
:tol 1.0E-6,
:quantile-probabilities q-probs,
:aggregation-depth 2,
:fit-intercept true,
:label-col "label",
:censor-col "censor",
:prediction-col "prediction",
:features-col "features"}
props (-> (merge defaults params))]
(interop/instantiate AFTSurvivalRegression props)))
(defn isotonic-regression [params]
(let [defaults {:prediction-col "prediction",
:features-col "features",
:isotonic true,
:label-col "label",
:feature-index 0}
props (-> (merge defaults params))]
(interop/instantiate IsotonicRegression props)))
(defn fm-regressor [params]
(let [defaults {:max-iter 100,
:step-size 1.0,
:tol 1.0E-6,
:reg-param 0.0,
:seed 891375198,
:mini-batch-fraction 1.0,
:fit-intercept true,
:label-col "label",
:factor-size 8,
:fit-linear true,
:prediction-col "prediction",
:init-std 0.01,
:features-col "features",
:solver "adamW"}
props (-> (merge defaults params))]
(interop/instantiate FMRegressor props)))
;; Docs
(docs/alter-docs-in-ns!
'zero-one.geni.ml.regression
[(-> docs/spark-docs :classes :ml :regression)])
;; Aliases
(import-fn generalized-linear-regression generalised-linear-regression)
(import-fn generalized-linear-regression glm)