CategoryDependencyGenerator Class Reference

Creates multipal features based on categorical values and statistical dependency strength by Age,Gender groups. More...

#include <FeatureGenerator.h>

Inheritance diagram for CategoryDependencyGenerator:

Public Member Functions
void	set_signal_ids (MedSignals &sigs)
void	init_tables (MedDictionarySections &dict)
void	init_defaults ()
virtual void	copy (FeatureGenerator *generator)
int	init (map< string, string > &mapper)
	The parsed fields from init command.
int	update (map< string, string > &mapper)
	Virtual to update object from parsed fields.
void	set_names ()
int	filter_features (unordered_set< string > &validFeatures)
	summary> prints summary of generator job.
int	_learn (MedPidRepository &rep, const MedSamples &samples, vector< RepProcessor * > processors)
int	_generate (PidDynamicRec &rec, MedFeatures &features, int index, int num, vector< float * > &_p_data)
int	nfeatures ()
void	get_required_signal_categories (unordered_map< string, vector< string > > &signal_categories_in_use) const
	returns for each used signal it's used categories
Public Member Functions inherited from FeatureGenerator
virtual void	prepare (MedFeatures &features, MedPidRepository &rep, MedSamples &samples)
virtual void	get_p_data (MedFeatures &features, vector< float * > &_p_data)
void	get_p_data (MedFeatures &features)
virtual void	clear ()
void	get_required_signal_names (unordered_set< string > &signalNames)
virtual void	set_required_signal_ids (MedDictionarySections &dict)
void	get_required_signal_ids (unordered_set< int > &signalIds)
virtual void	get_generated_features (unordered_set< string > &names_list)
virtual void	fit_for_repository (MedPidRepository &rep)
	Prepartion and adjustment for model based on repository.
int	learn (MedPidRepository &rep, const MedSamples &samples, vector< RepProcessor * > processors)
int	learn (MedPidRepository &rep, const MedSamples &samples)
int	_generate (PidDynamicRec &in_rep, MedFeatures &features, int index, int num)
int	generate (PidDynamicRec &in_rep, MedFeatures &features, int index, int num)
int	generate (PidDynamicRec &in_rep, MedFeatures &features)
int	generate (MedPidRepository &rep, int id, MedFeatures &features)
int	generate (MedPidRepository &rep, int id, MedFeatures &features, int index, int num)
virtual int	_generate (MedFeatures &features)
int	generate (MedFeatures &features)
virtual int	init (void *generator_params)
virtual void	make_summary ()
void *	new_polymorphic (string derived_class_name)
	for polymorphic classes that want to be able to serialize/deserialize a pointer * to the derived class given its type one needs to implement this function to return a new to the derived class given its type (as in my_type)
size_t	get_generator_size ()
size_t	generator_serialize (unsigned char *blob)
virtual void	print ()
virtual void	dprint (const string &pref, int fg_flag)
Public Member Functions inherited from SerializableObject
virtual int	version () const
	Relevant for serializations.
virtual string	my_class_name () const
	For better handling of serializations it is highly recommended that each SerializableObject inheriting class will implement the next method.
virtual void	serialized_fields_name (vector< string > &field_names) const
	The names of the serialized fields.
virtual void	pre_serialization ()
virtual void	post_deserialization ()
virtual size_t	get_size ()
	Gets bytes sizes for serializations.
virtual size_t	serialize (unsigned char *blob)
	Serialiazing object to blob memory. return number ob bytes wrote to memory.
virtual size_t	deserialize (unsigned char *blob)
	Deserialiazing blob to object. returns number of bytes read.
size_t	serialize_vec (vector< unsigned char > &blob)
size_t	deserialize_vec (vector< unsigned char > &blob)
virtual size_t	serialize (vector< unsigned char > &blob)
virtual size_t	deserialize (vector< unsigned char > &blob)
virtual int	read_from_file (const string &fname)
	read and deserialize model
virtual int	write_to_file (const string &fname)
	serialize model and write to file
virtual int	read_from_file_unsafe (const string &fname)
	read and deserialize model without checking version number - unsafe read
int	init_from_string (string init_string)
	Init from string.
int	init_params_from_file (string init_file)
int	init_param_from_file (string file_str, string &param)
int	update_from_string (const string &init_string)
virtual string	object_json () const

Data Fields
string	signalName
	the signal name
int	signalId
int	time_channel
	n >= 0 : use time channel n , default: 0.
int	val_channel
	n >= 0 : use val channel n , default : 0.
int	win_from
	time window for feature: win_from is the minimal time before from the prediction time
int	win_to
	time window for feature: win_to is the maximal time before the prediction time
int	time_unit_win
	the time unit in which the windows are given. Default: Undefined
int	min_age
	minimal age for testing statistical dependency
int	max_age
	maximal age for testing statistical dependency
int	age_bin
	age bin for testing statistical dependency
string	regex_filter
	regex filter for filtering categories in learn
string	remove_regex_filter
	remove regex filter for filtering categories in learn
int	min_code_cnt
	minimal number of occourences to consider signal
float	fdr
	the FDR value
int	take_top
	maximal number of features to create
float	lift_below
	filter lift to keep below it
float	lift_above
	filter lift to keep above it
float	filter_child_pval_diff
	below this threshold of pvalue diff change to remove child category (with AND condition on average lift change)
float	filter_child_lift_ratio
	below this threshold of lift change to remove child category
float	filter_child_count_ratio
	If child ratio count is too similar, small change from parent code - keep only paretn code.
float	filter_child_removed_ratio
	If child removed ratio is beyond this and has other child taken - remove parent.
category_stat_test	stat_metric
	statistical test
float	chi_square_at_least
	chi_square arg to test for at least that change in lift to measure bigger diffrence
int	minimal_chi_cnt
	chi_square arg to keep at least count to use row in calc
int	sort_by_chi = 0
	sort results by chi-square
int	max_depth
	maximal depth to go in heirarchy
int	max_parents
	controls maximum parents count
bool	use_fixed_lift
	If true will also sort be lifts below 1.
bool	filter_hierarchy
bool	verbose
	Apply hierarchy filtering.
bool	verbose_full
	If true will print a lot - table of all stats for each code.
string	verbose_full_file
	output file for verbose_full debug in learn
string	feature_prefix
	additional prefix to add to name to describe the feature
bool	generate_with_counts
	If true will generate feature with counts not just as set.
vector< vector< string > >	filter_set_by_val_channel
	filter set by value channels. can be initialized by "filter_set_by_val_channel_X":"string_set_for_val_channel_X",
vector< string >	filter_set_by_val_channel_names
	naming for each set matched filter_set_by_val_channel variable
float	male_regression_cntrl_lower
	lower limit mask on outcome for controls - important inregression
float	male_regression_cntrl_upper
	upper limit mask on outcome for controls - important inregression
float	male_regression_case_lower
	lower limit mask on outcome for cases - important inregression
float	male_regression_case_upper
	upper limit mask on outcome for cases - important inregression
float	female_regression_cntrl_lower
	lower limit mask on outcome for controls - important inregression
float	female_regression_cntrl_upper
	upper limit mask on outcome for controls - important inregression
float	female_regression_case_lower
	lower limit mask on outcome for cases - important inregression
float	female_regression_case_upper
	upper limit mask on outcome for cases - important inregression
Data Fields inherited from FeatureGenerator
FeatureGeneratorTypes	generator_type = FTR_GEN_LAST
	Type.
vector< string >	names
	Feature name.
int	learn_nthreads = 16
int	pred_nthreads = 16
float	missing_val = (float)MED_MAT_MISSING_VALUE
	Missing value.
vector< string >	tags
	Tags - for defining labels or groups. may be used later for filtering for example.
int	iGenerateWeights = 0
	Feature/Weights generator.
vector< float * >	p_data
vector< string >	req_signals
vector< int >	req_signal_ids
int	serial_id

Additional Inherited Members
Static Public Member Functions inherited from FeatureGenerator
static FeatureGenerator *	make_generator (string name)
static FeatureGenerator *	make_generator (string name, string params)
static FeatureGenerator *	make_generator (FeatureGeneratorTypes type)
static FeatureGenerator *	make_generator (FeatureGeneratorTypes type, string params)
static FeatureGenerator *	create_generator (string &params)

Detailed Description

Creates multipal features based on categorical values and statistical dependency strength by Age,Gender groups.

Member Function Documentation

_generate()

int CategoryDependencyGenerator::_generate ( PidDynamicRec &  rec, MedFeatures &  features, int  index, int  num, vector< float * > &  _p_data  )

virtual

Reimplemented from FeatureGenerator.

_learn()

int CategoryDependencyGenerator::_learn ( MedPidRepository &  rep, const MedSamples &  samples, vector< RepProcessor * >  processors  )

virtual

Reimplemented from FeatureGenerator.

copy()

virtual void CategoryDependencyGenerator::copy ( FeatureGenerator *  generator )

inlinevirtual

Reimplemented from FeatureGenerator.

filter_features()

int CategoryDependencyGenerator::filter_features ( unordered_set< string > &  validFeatures )

virtual

summary> prints summary of generator job.

optional, called after generate. for example - prints how many values were missing value /summary>

Reimplemented from FeatureGenerator.

get_required_signal_categories()

void CategoryDependencyGenerator::get_required_signal_categories ( unordered_map< string, vector< string > > &  signal_categories_in_use ) const

virtual

returns for each used signal it's used categories

Reimplemented from FeatureGenerator.

init()

int CategoryDependencyGenerator::init ( map< string, string > &  mapper )

virtual

The parsed fields from init command.

        if (it->first == "signal")
            signalName = it->second;
        else if (it->first == "val_channel")
            val_channel = med_stoi(it->second);
        else if (it->first == "time_channel")
            time_channel = med_stoi(it->second);
        else if (it->first == "win_from")
            win_from = med_stoi(it->second);
        else if (it->first == "win_to")
            win_to = med_stoi(it->second);
        else if (it->first == "time_unit_win")
            time_unit_win = med_time_converter.string_to_type(it->second);
        else if (it->first == "regex_filter")
            regex_filter = it->second;
        else if (it->first == "remove_regex_filter")
            remove_regex_filter = it->second;
        else if (it->first == "min_age")
            min_age = med_stoi(it->second);
        else if (it->first == "max_age")
            max_age = med_stoi(it->second);
        else if (it->first == "age_bin")
            age_bin = med_stoi(it->second);
        else if (it->first == "min_code_cnt")
            min_code_cnt = med_stoi(it->second);
        else if (it->first == "fdr")
            fdr = med_stof(it->second);
        else if (it->first == "take_top")
            take_top = med_stoi(it->second);
        else if (it->first == "filter_hierarchy")
            filter_hierarchy = med_stoi(it->second) > 0;
        else if (it->first == "lift_below")
            lift_below = med_stof(it->second);
        else if (it->first == "lift_above")
            lift_above = med_stof(it->second);
        else if (it->first == "filter_child_count_ratio")
            filter_child_count_ratio = med_stof(it->second);
        else if (it->first == "filter_child_lift_ratio")
            filter_child_lift_ratio = med_stof(it->second);
        else if (it->first == "filter_child_pval_diff")
            filter_child_pval_diff = med_stof(it->second);
        else if (it->first == "filter_child_removed_ratio")
            filter_child_removed_ratio = med_stof(it->second);
        else if (it->first == "chi_square_at_least")
            chi_square_at_least = med_stof(it->second);
        else if (it->first == "sort_by_chi")
            sort_by_chi = med_stoi(it->second);
        else if (it->first == "minimal_chi_cnt")
            minimal_chi_cnt = med_stoi(it->second);
        else if (it->first == "use_fixed_lift")
            use_fixed_lift = med_stoi(it->second) > 0;
        else if (it->first == "verbose")
            verbose = med_stoi(it->second) > 0;
        else if (it->first == "verbose_full")
            verbose_full = med_stoi(it->second) > 0;
        else if (it->first == "verbose_full_file")
            verbose_full_file = it->second;
        else if (it->first == "feature_prefix")
            feature_prefix = it->second;
        else if (it->first == "stat_metric") {
            if (conv_map_stats.find(it->second) != conv_map_stats.end())
                stat_metric = category_stat_test(conv_map_stats.at(it->second));
            else
                MTHROW_AND_ERR("Unknown stat_test \"%s\". options are: %s\n",
                    it->second.c_str(), medial::io::get_list(conv_map_stats).c_str());
        }
        else if (it->first == "max_depth")
            max_depth = med_stoi(it->second);
        else if (it->first == "max_parents")
            max_parents = med_stoi(it->second);
        else if (it->first == "generate_with_counts")
            generate_with_counts = med_stoi(it->second) > 0;
        else if (it->first == "regression_cntrl_lower") {
            male_regression_cntrl_lower = med_stoi(it->second);
            female_regression_cntrl_lower = med_stoi(it->second);
        }
        else if (it->first == "regression_cntrl_upper") {
            male_regression_cntrl_upper = med_stoi(it->second);
            female_regression_cntrl_upper = med_stoi(it->second);
        }
        else if (it->first == "regression_case_lower") {
            male_regression_case_lower = med_stoi(it->second);
            female_regression_case_lower = med_stoi(it->second);
        }
        else if (it->first == "regression_case_upper") {
            male_regression_case_upper = med_stoi(it->second);
            female_regression_case_upper = med_stoi(it->second);
        }
        else if (it->first == "male_regression_cntrl_lower")
            male_regression_cntrl_lower = med_stoi(it->second);
        else if (it->first == "male_regression_cntrl_upper")
            male_regression_cntrl_upper = med_stoi(it->second);
        else if (it->first == "male_regression_case_lower")
            male_regression_case_lower = med_stoi(it->second);
        else if (it->first == "male_regression_case_upper")
            male_regression_case_upper = med_stoi(it->second);
        else if (it->first == "female_regression_cntrl_lower")
            female_regression_cntrl_lower = med_stoi(it->second);
        else if (it->first == "female_regression_cntrl_upper")
            female_regression_cntrl_upper = med_stoi(it->second);
        else if (it->first == "female_regression_case_lower")
            female_regression_case_lower = med_stoi(it->second);
        else if (it->first == "female_regression_case_upper")
            female_regression_case_upper = med_stoi(it->second);
        else if (boost::starts_with(it->first, prefix_str)) {
            int val_channel_f = med_stoi(it->first.substr(prefix_str.length()));
            if (filter_set_by_val_channel.size() <= val_channel_f)
                filter_set_by_val_channel.resize(val_channel_f + 1);
            vector<string> &f_v_sets = filter_set_by_val_channel[val_channel_f];
            boost::split(f_v_sets, it->second, boost::is_any_of(","));
        }
        else if (boost::starts_with(it->first, prefix_str_names)) {
            int val_channel_f = med_stoi(it->first.substr(prefix_str_names.length()));
            if (filter_set_by_val_channel_names.size() <= val_channel_f)
                filter_set_by_val_channel_names.resize(val_channel_f + 1);
            filter_set_by_val_channel_names[val_channel_f] = it->second;
        }
        else if (it->first == "fg_type") {}
        else if (it->first == "tags") { boost::split(tags, it->second, boost::is_any_of(",")); }
        else
            MTHROW_AND_ERR("Unknown parameter \'%s\' for CategoryDependencyGenerator\n", it->first.c_str())

[CategoryDependencyGenerator::init]

[CategoryDependencyGenerator::init]

Reimplemented from FeatureGenerator.

init_defaults()

void CategoryDependencyGenerator::init_defaults ( )

virtual

Reimplemented from FeatureGenerator.

init_tables()

void CategoryDependencyGenerator::init_tables ( MedDictionarySections &  dict )

virtual

Reimplemented from FeatureGenerator.

nfeatures()

int CategoryDependencyGenerator::nfeatures ( )

virtual

Reimplemented from FeatureGenerator.

set_names()

void CategoryDependencyGenerator::set_names ( )

virtual

Reimplemented from FeatureGenerator.

set_signal_ids()

void CategoryDependencyGenerator::set_signal_ids ( MedSignals &  sigs )

virtual

Reimplemented from FeatureGenerator.

update()

int CategoryDependencyGenerator::update ( map< string, string > &  map )

virtual

Virtual to update object from parsed fields.

Reimplemented from SerializableObject.

Field Documentation

verbose

bool CategoryDependencyGenerator::verbose

Apply hierarchy filtering.

in Learn will print selected features

---

The documentation for this class was generated from the following files:

• Internal/MedProcessTools/MedProcessTools/FeatureGenerator.h
• Internal/MedProcessTools/MedProcessTools/CategoryDependencyGenerator.cpp