B >?[1@sdZddlmZddlZddlZddlZddlZddlZddlZddl m Z ddl m Z m Z ddlmZddlmZmZddlmZdad d d d d gZdddZddZGdddeZGdddZedkrddlmZmZddZeeeZ dS)z; Classifiers that make use of the external 'Weka' package. )print_functionN)stdin) integer_types string_types)DictionaryProbDist)java config_java) ClassifierI.z/usr/share/wekaz/usr/local/share/wekaz /usr/lib/wekaz/usr/local/lib/wekacCst|dk r|atdkrt}dtjkr:|dtjdxb|D]Z}tjtj|dr@tj|dat t}|rt dt|fn t dtt tq@Wtdkrt ddS)NZWEKAHOMErzweka.jarz[Found Weka: %s (version %s)]z[Found Weka: %s]zUnable to find weka.jar! Use config_weka() or set the WEKAHOME environment variable. For more information about Weka, please see http://www.cs.waikato.ac.nz/ml/weka/) r_weka_classpath _weka_searchosenvironinsertpathexistsjoin_check_weka_versionprint LookupError) classpathZ searchpathrversionr1lib/python3.7/site-packages/nltk/classify/weka.py config_weka%s&    rc Csfyt|}Wn"ttfk r(Yn dSz$y |dStk rRdSXWd|XdS)Nzweka/core/version.txt)zipfileZZipFile SystemExitKeyboardInterruptreadKeyErrorclose)ZjarZzfrrrrHs  rc@sbeZdZddZddZddZddZd d Zd d Zd ddddddZ e dgdfddZ dS)WekaClassifiercCs||_||_dS)N) _formatter_model)self formattermodel_filenamerrr__init__YszWekaClassifier.__init__cCs||dddgS)Nz-p0z -distribution)_classify_many)r$ featuresetsrrrprob_classify_many]sz!WekaClassifier.prob_classify_manycCs||ddgS)Nz-pr()r))r$r*rrr classify_many`szWekaClassifier.classify_manyc Cstt}ztj|d}|j||dd|jd|g|}t |t t j t j d\}}|r~|s~d|krrt dn t d|||tjd Sx&t|D]}ttj||qWt|XdS) Nz test.arffz!weka.classifiers.bayes.NaiveBayesz-lz-T)rstdoutstderrzIllegal options: -distributionzOThe installed version of weka does not support probability distribution output.z"Weka failed to generate output: %s )rtempfilemkdtempr rrr"writer#rr subprocessPIPE ValueErrorparse_weka_outputdecoderencodingsplitlistdirremovermdir) r$r*optionstemp_dirZ test_filenamecmdr-r.frrrr)cs2 zWekaClassifier._classify_manycCs2ddtd|D}tt|j|}t|S)NcSsg|]}|rt|qSr)stripfloat).0vrrr sz:WekaClassifier.parse_weka_distribution..z[*,]+)rer9dictzipr"labelsr)r$sZprobsrrrparse_weka_distributionsz&WekaClassifier.parse_weka_distributioncsx0t|D]$\}}|dr ||d}Pq W|ddddddgkrbdd|d dDS|dddddd gkrfd d|d dDStd |drd d|DSx|ddD] }t|qWtd|ddS)Nzinst#rZactualZ predictederrorZ predictioncSs*g|]"}|r|dddqS):)rAr9)rClinerrrrEsz4WekaClassifier.parse_weka_output..rOZ distributioncs&g|]}|r|dqS))rArKr9)rCrP)r$rrrEsz^0 \w+ [01]\.[0-9]* \?\s*$cSs g|]}|r|dqS)rO)rAr9)rCrPrrrrEs zRUnhandled output format -- your version of weka may not be supported. Header: %s) enumeraterA startswithr9rFmatchrr5)r$linesirPr)r$rr6s*    z WekaClassifier.parse_weka_outputz!weka.classifiers.bayes.NaiveBayeszweka.classifiers.trees.J48z#weka.classifiers.functions.Logisticzweka.classifiers.functions.SMOzweka.classifiers.lazy.KStarzweka.classifiers.rules.JRip) naivebayeszC4.5Zlog_regressionZsvmZkstarZripperrXTc Cstt|}t}ztj|d}|||||j krJ|j |} n ||j kr^|} n t d|| d|d|g} | t |7} |rt j} nd} t| t| dt||Sx&t|D]} ttj|| qWt|XdS)Nz train.arffzUnknown classifier %sz-dz-t)rr-)rARFF_Formatter from_trainr0r1r rrr2_CLASSIFIER_CLASSvaluesr5listr3r4rr r!r:r;r<) clsr&r* classifierr=quietr%r>Ztrain_filenameZ javaclassr?r-r@rrrtrains*        zWekaClassifier.trainN) __name__ __module__ __qualname__r'r+r,r)rKr6r[ classmethodrarrrrr!Xs ,1r!c@sVeZdZdZddZddZddZdd Zed d Z d d Z dddZ ddZ dS)rYz Converts featuresets and labeled featuresets to ARFF-formatted strings, appropriate for input into Weka. Features and classes can be specified manually in the constructor, or may be determined from data using ``from_train``. cCs||_||_dS)a) :param labels: A list of all class labels that can be generated. :param features: A list of feature specifications, where each feature specification is a tuple (fname, ftype); and ftype is an ARFF type string such as NUMERIC or STRING. N)_labels _features)r$rIfeaturesrrrr'szARFF_Formatter.__init__cCs|||S)zBReturns a string representation of ARFF output for the given data.)header_section data_section)r$tokensrrrformatszARFF_Formatter.formatcCs t|jS)zReturns the list of classes.)r]rf)r$rrrrIszARFF_Formatter.labelscCs0t|dst|d}||||dS)z.Writes ARFF data to a file for the given data.r2wN)hasattropenr2rlr )r$Zoutfilerkrrrr2s  zARFF_Formatter.writecCstdd|D}i}x|D]\}}x|D]\}}tt|trJd}nFtt|tttfrdd}n,tt|trxd}n|dkrq.n td|| |||krtd||||<q.WqWt |}t ||S) z Constructs an ARFF_Formatter instance with class labels and feature types determined from the given data. Handles boolean, numeric and string (note: not nominal) types. css|]\}}|VqdS)Nr)rCtoklabelrrr &sz,ARFF_Formatter.from_train..z {True, False}ZNUMERICSTRINGNzUnsupported value type %rzInconsistent type for %s) setitems issubclasstypeboolrrBrr5getsortedrY)rkrIrhrprqfnamefvalftyperrrrZs$   zARFF_Formatter.from_traincCsXddt}|d7}x"|jD]\}}|d||f7}q W|ddd|jf7}|S)z#Returns an ARFF header as a string.z3% Weka ARFF file % Generated automatically by NLTK z%% %s z@RELATION rel z@ATTRIBUTE %-30r %s z@ATTRIBUTE %-30r {%s} z-label-,)timeZctimergrrf)r$rJr{r}rrrri>s zARFF_Formatter.header_sectionNc Cs|dkr|ot|dttf}|s0dd|D}d}xN|D]F\}}x*|jD] \}}|d|||7}qJW|d||7}q:W|S)a Returns the ARFF data section for the given data. :param tokens: a list of featuresets (dicts) or labelled featuresets which are tuples (featureset, label). :param labeled: Indicates whether the given tokens are labeled or not. If None, then the tokens will be assumed to be labeled if the first token's value is a tuple or list. NrcSsg|] }|dfqS)Nr)rCrprrrrEbsz/ARFF_Formatter.data_section..z @DATA z%s,z%s ) isinstancetupler]rg _fmt_arff_valry)r$rkZlabeledrJrprqr{r}rrrrjSs zARFF_Formatter.data_sectioncCs@|dkr dSt|ttfr"d|St|tr4d|Sd|SdS)N?z%sz%r)rrxrrB)r$r|rrrrms zARFF_Formatter._fmt_arff_val)N) rbrcrd__doc__r'rlrIr2 staticmethodrZrirjrrrrrrYs   rY__main__) names_demobinary_names_demo_featurescCstd|dS)Nz/tmp/name.modelzC4.5)r!ra)r*rrrmake_classifier{sr)N)!rZ __future__rrr0r r3rFrsysrZsixrrZnltk.probabilityrZnltk.internalsrrZnltk.classify.apir r r rrr!rYrbZnltk.classify.utilrrrr_rrrr s6     #$}