fitGWP: Fit a Generalized Word Power model.

In keblu/GWP: Generalized Word Power Approach

Description

Fit a Generalized Word Power model on a target variable using word frequency data with an elastic-net model.

Usage

fitGWP(frequencies, responseData, alpha = c(0, 0.2, 0.4, 0.6, 0.8, 1),
  lowerLimit = 0.2)

Arguments

frequencies	frequencies data.table generated using the function computeFrequencies.
responseData	data.frame with column regID and y. regID should match with the regID of the frequencies input while y is the response variable.
alpha	alpha parameter for an elastic-net model.
lowerLimit	lower limit on the number of time a sentiment word must appear over all the period. If the value is above or equal to 1, this value is absolute. If the value is below 1, this value is in percentage term.

Value

A list with the following elements:

• scores: data.frame wiht the column word, score, and importance. the word column correspond to the sentiment words, the score column correspond to the polarity of the sentiment words, and the importance column correspond to the importance of each word withing the calbration sample.

• sentiment: data.frame with colmn regID, SentimentScores, and y. the column regID correspond to the regID of each observation, the columne SentimentScores is the sentiment for that observation, and the column y is the initially given reponse variable.

Examples

# Load example data
data("corpus",  package = "GWP")
data("vix",  package = "GWP")
# Setup the lexicons
sentimentWord <- sentometrics::list_lexicons$LM_en$x
shifterWord <- sentometrics::list_valence_shifters$en[, c("x", "y")]

# Generate the frequency data
frequencies <- computeFrequencies(corpus, sentimentWord, shifterWord, clusterSize = 5)

# Calibrate the lexicon
res = fitGWP(frequencies = frequencies, responseData = vix)

keblu/GWP documentation built on June 15, 2020, 11:43 p.m.