Documentation
¶
Index ¶
- Variables
- func EulerDistance(X, Y, W []float64) float64
- func HuffmanDistance(X, Y, W []float64) float64
- type Class
- type DistanceMethodType
- type IntentClassifier
- type KNearestNeighbors
- type NaiveBayes
- func (nb *NaiveBayes) Classify(text string) (string, float64)
- func (nb *NaiveBayes) GobDecode(data []byte) error
- func (nb *NaiveBayes) GobEncode() ([]byte, error)
- func (nb *NaiveBayes) Learn(text, class string)
- func (nb *NaiveBayes) LearnDocument(document types.Document, class string)
- func (nb *NaiveBayes) LearnSentence(sentence types.Sentence, class string)
- func (nb *NaiveBayes) Load(filePath string) error
- func (nb *NaiveBayes) Save(file string) error
- type SortedDistance
Constants ¶
This section is empty.
Variables ¶
var ( // ErrNotClassified indicates that a document could not be classified ErrNotClassified = errors.New("unable to classify document") )
Functions ¶
Types ¶
type Class ¶
type Class struct {
Name string
Counter int
Words map[string]types.Word
Probability int
Temp_tokenProbabilities float64
}
*
- The possible class outcomes.
type DistanceMethodType ¶
type DistanceMethodType uint8
const ( //Euler distance method DMT_EulerMethod DistanceMethodType = iota // Huffman DMT_HuffmanMethod )
type IntentClassifier ¶
type IntentClassifier struct {
Feat2cat map[string]map[string]int
CatCount map[string]int
Mu sync.RWMutex
Tokenizer string
}
func NewIntentClassifier ¶
func NewIntentClassifier() *IntentClassifier
New initializes a new naive Classifier using the standard tokenizer
func (*IntentClassifier) Classify ¶
func (c *IntentClassifier) Classify(r string) (string, error)
Classify attempts to classify a document. If the document cannot be classified (eg. because the classifier has not been trained), an error is returned.
func (*IntentClassifier) Load ¶
func (c *IntentClassifier) Load(filePath string) error
Load from the output file.
type KNearestNeighbors ¶
type KNearestNeighbors struct {
/**
* The number of neighbors to consider when making a prediction.
*
* @var int
*/
K int
/**
* The distance function to use when computing the distances.
*
*/
DistanceMethod DistanceMethodType
/**
* Should we use the inverse distances as confidence scores when
* making predictions?
* Weight define the weight vector for multi-dimension data
*/
Weight []float64
// The training samples that make up the neighborhood of the problem space.
Samples [][]float64
//The memoized labels of the training set.
Labels []string
}
*
K Nearest Neighbors *
A distance-based algorithm that locates the K nearest neighbors from the
training set and uses a majority vote to classify the unknown sample. K
Nearest Neighbors is considered a lazy learning Estimator because it does all
of its computation at prediction time. *
@category Machine Learning
@author Bruce Mubangwa
**usage
train := [][]float64{
[]float64{5.3, 3.7},
[]float64{5.1, 3.8},
[]float64{7.2, 3},
[]float64{5.4, 3.4},
[]float64{5.1, 3.3},
[]float64{5.4, 3.9},
[]float64{7.4, 2.8},
[]float64{6.1, 2.8},
[]float64{7.3, 2.9},
[]float64{6, 2.7},
[]float64{5.8, 2.8},
[]float64{6.3, 2.3},
[]float64{5.1, 2.5},
[]float64{6.3, 2.5},
[]float64{5.5, 2.4},
}
labels := []string{
"Setosa",
"Setosa",
"Virginica",
"Setosa",
"Setosa",
"Setosa",
"Virginica",
"Versicolor",
"Virginica",
"Versicolor",
"Virginica",
"Versicolor",
"Versicolor",
"Versicolor",
"Versicolor",
}
dm := classifiers.DMT_EulerMethod
w := []float64{0.5, 0.5}
var test [][]float64
test = [][]float64{
[]float64{5.2, 3.1},
}
for i := 1; i < 12; i++ {
knn := classifiers.NewKNearestNeighbors(i, dm, w)
knn.LearnBatch(train, labels)
res := knn.Classify(test)
if res[0] != "Setosa" {
fmt.Printf("k = %d failed", i)
fmt.Println()
}
}
func NewKNearestNeighbors ¶
func NewKNearestNeighbors(k int, dm DistanceMethodType, w []float64) *KNearestNeighbors
func (*KNearestNeighbors) Classify ¶
func (k *KNearestNeighbors) Classify(test [][]float64) []string
*
- Make predictions from a dataset.
- Classify, train is an nxp matrix, where n denotes the n training data,
- and p represents the number of attributes. The test is an mxp matrix.
func (*KNearestNeighbors) LearnBatch ¶
func (k *KNearestNeighbors) LearnBatch(train [][]float64, label []string)
*
- Store the sample and outcome arrays. No other work to be done as this is
- a lazy learning algorithm. *
func (*KNearestNeighbors) Load ¶
func (k *KNearestNeighbors) Load(filePath string) error
Load from the output file.
type NaiveBayes ¶
type NaiveBayes struct {
// contains filtered or unexported fields
}
`bayes`: A Naive-Bayes classifier for Go ¶
`bayes` takes a document (piece of text), and tells you what category that document belongs to.
## What can I use this for?
You can use this for categorizing any text content into any arbitrary set of **categories**. For example:
- is an email **spam**, or **not spam** ? - is a news article about **technology**, **politics**, or **sports** ? - is a piece of text expressing **positive** emotions, or **negative** emotions?
## Usage
classifier := classifiers.NewNaiveBayes()
classifier.Learn("amazing, awesome movie!! Yeah!! Oh boy.", "positive")
classifier.Learn("Sweet, this is incredibly, amazing, perfect, great!!", "positive")
classifier.Learn("terrible, shitty thing. Damn. Sucks!!", "negative")
fmt.Println(classifier.Classify("awesome, cool shitty thing"))
func NewNaiveBayes ¶
func NewNaiveBayes() *NaiveBayes
func (*NaiveBayes) Classify ¶
func (nb *NaiveBayes) Classify(text string) (string, float64)
*
- Determine what category or class `text` belongs to.
func (*NaiveBayes) GobDecode ¶
func (nb *NaiveBayes) GobDecode(data []byte) error
GobDecode implements GoDecoder.
func (*NaiveBayes) GobEncode ¶
func (nb *NaiveBayes) GobEncode() ([]byte, error)
GobEncode implements GobEncoder. This is necessary because RNN contains several unexported fields. It would be easier to simply export them by changing to uppercase, but for comparison purposes, I wanted to keep the field names the same between Go and the original Python code.
func (*NaiveBayes) Learn ¶
func (nb *NaiveBayes) Learn(text, class string)
*
- train our naive-bayes classifier by telling it what `category`
- the `text` corresponds to.
func (*NaiveBayes) LearnDocument ¶
func (nb *NaiveBayes) LearnDocument(document types.Document, class string)
func (*NaiveBayes) LearnSentence ¶
func (nb *NaiveBayes) LearnSentence(sentence types.Sentence, class string)
func (*NaiveBayes) Load ¶
func (nb *NaiveBayes) Load(filePath string) error
Load from the output file.
func (*NaiveBayes) Save ¶
func (nb *NaiveBayes) Save(file string) error
type SortedDistance ¶
SortedDistance
func NewSortedDistance ¶
func NewSortedDistance(size int) *SortedDistance
NewSortedDistance initial the SortedDistance with the size
func (*SortedDistance) Less ¶
func (s *SortedDistance) Less(i, j int) bool
Less return true if [i] < [j].
func (*SortedDistance) SelectTopKIdx ¶
func (s *SortedDistance) SelectTopKIdx(k int) []int
SelectTopKIdx return the index of the train
Source Files