[Julia]Naive Bayes using Julia

相似文件 换一批

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

1. # Method of computation
   
2. # Assumes it works over 2D arrays
   
3. 
   
4. function count_inputs_per_class(Y, C)
   
5.   # Don't know how to specify type of C
   
6.   # Anyhow since it's not known at compile time,
   
7.   # doesn't make a performance difference I suppose
   
8.   #
   
9.   # Maybe find a more fancy comprehension way
   
10.   # to do it one line
    
11.   N_c = Dict{Any, Int64}()
    
12.   for y in Y
    
13.     N_c[y] = haskey(N_c, y) ? N_c[y] + 1 : 1
    
14.   end
    
15.   N_c
    
16. end
    
17. 
    
18. # X - training set inputs
    
19. # Y - training set results
    
20. # c - class for which tokens qualify
    
21. # t - token for which count needs to be taken
    
22. function count_tokens_T_in_class(X, Y, c, t)
    
23.   count = 0
    
24.   for i = 1:size(X, 1)
    
25.     if !(Y[i] == c)
    
26.       continue
    
27.     end
    
28.     for x in X[i,:]
    
29.       if x == t
    
30.         count += 1
    
31.       end
    
32.     end
    
33.   end #Loop over training set
    
34.   count
    
35. end
    
36. 
    
37. function count_tokens_in_class(X, Y, c)
    
38.   count = 0
    
39.   for i = 1:size(X, 1)
    
40.     if (Y[i] == c)
    
41.       count += length(X[i,:])
    
42.     end
    
43.   end #Loop over training set
    
44.   count
    
45. end
    
46. 
    
47. # Currently supports only Multinomial Naive Bayes
    
48. # TODO: Add optional options hash as 3rd parameter
    
49. #
    
50. # X - Array containing training set inputs.
    
51. # Y - Vector containing training set results. (TODO: Change to vector parameter)
    
52. # returns - learned classification function gamma
    
53. #
    
54. function naive_bayes_fit(X::Array, Y::Array)
    
55.   V = unique(X)
    
56.   C = unique(Y)
    
57. 
    
58.   N = size(X, 1)
    
59.   N_c = count_inputs_per_class(Y, C)
    
60. 
    
61.   prior = Dict{Any, Float64}()
    
62.   #P_T_c = Array(Float64, (length(V), length(C)))
    
63.   # Might slow down due to the absence of type
    
64.   P_T_c = Dict()
    
65.   for c in C
    
66.     prior[c] = N_c[c] / N
    
67.     sum_T_ct = count_tokens_in_class(X, Y, c) + length(V)
    
68.     for t in V
    
69.       T_ct = count_tokens_T_in_class(X, Y, c, t) + 1
    
70.       P_T_c[(t, c)] = T_ct / sum_T_ct
    
71.     end
    
72.   end
    
73. 
    
74.   gamma(d) = naive_bayes_predict(prior, P_T_c, C, d)
    
75. end
    
76. 
    
77. function naive_bayes_predict(prior, P_T_c, C, d)
    
78.   n_classes = length(C)
    
79.   scores = Array(Float64, n_classes)
    
80.   for i = 1:n_classes; c = C[i]
    
81.     scores[i] = prior[c]
    
82.     for t in d
    
83.       if (haskey(P_T_c, (t, c)))
    
84.         scores[i] += log(P_T_c[(t, c)])
    
85.       end
    
86.     end
    
87.   end
    
88.   C[indmax(scores)]
    
89. end

复制代码

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏0 回帖

关键词：naive Using Bayes Julia baye difference training suppose

1. X = reshape(rand(0:5, 600), 6, 100)
   
2. Y = [1:6]
   
3. 
   
4. gamma = naive_bayes_fit(X, Y)
   
5. print(gamma(X[1]))

复制代码