Weighted Voting Aggregation for Multi-Voter Classification

Aggregates classification votes from multiple voters using weighted scoring. Each voter assigns one of four labels ("Positive", "Negative", "Neutral", "Other") to each item, and votes are combined using user-specified weights. The "Other" category contributes half-weight to both "Negative" and "Neutral" categories.

Usage

WeightedVote(
  vote_data = data.frame(),
  weights = double(),
  ties.method = c("random", "first", "last")
)

Arguments

vote_data

A data frame or matrix where:

• Rows represent items to be classified

• Columns represent individual voters

• Cell values must be one of: "Positive", "Negative", "Neutral", "Other"

Row names (if present) are preserved in the output.

weights

Named numeric vector specifying the weight for each voter. Names must exactly match colnames(vote_data). Weights should be non-negative, with higher values indicating greater influence.

ties.method

Character string specifying how to break ties. Must be one of "random", "first", or "last".

Value

Character vector of final aggregated labels, one per row in vote_data. Each element is one of "Positive", "Negative", or "Neutral" (the winner with highest weighted score). If input has row names, they are preserved as names in the output vector.

Scoring Logic

For each item (row), scores are computed as:

Positive score

Sum of weights where vote = "Positive"

Negative score

Sum of weights where vote = "Negative" + 0.5 × sum of weights where vote = "Other"

Neutral score

Sum of weights where vote = "Neutral" + 0.5 × sum of weights where vote = "Other"

The category with the highest score wins. In case of ties, the first category (in order: Positive, Negative, Neutral) is selected.

See also

table for vote counting, max.col for winner selection

Examples

if (FALSE) { # \dontrun{
# Example 1: Simple majority voting
vote_data <- data.frame(
  Expert1 = c("Positive", "Negative", "Neutral", "Positive"),
  Expert2 = c("Positive", "Neutral", "Other", "Negative"),
  Expert3 = c("Negative", "Positive", "Neutral", "Positive"),
  row.names = c("Gene1", "Gene2", "Gene3", "Gene4")
)

weights <- c(Expert1 = 1, Expert2 = 1, Expert3 = 1)
result <- WeightedVote(vote_data, weights)
# Gene1: Positive(2) vs Negative(1) → "Positive"
# Gene2: each has 1 vote → "Positive" (first in tie)
# Gene3: Neutral(1.5) vs Other(0.5) → "Neutral"
# Gene4: Positive(2) vs Negative(1) → "Positive"

# Example 2: Unequal weights (expert confidence levels)
weights <- c(Expert1 = 3, Expert2 = 1, Expert3 = 2)
result <- WeightedVote(vote_data, weights)

# Example 3: "Other" category splits weight
vote_data <- data.frame(
  Voter1 = c("Positive", "Other", "Negative"),
  Voter2 = c("Other", "Neutral", "Other")
)
weights <- c(Voter1 = 1, Voter2 = 1)
result <- WeightedVote(vote_data, weights)
# Row1: Positive(1) + Other(0.5 to Neg/Neu) → Positive wins
# Row2: Other(0.5 to Neg/Neu) + Neutral(1) → Neutral wins
# Row3: Negative(1) + Other(0.5 to Neg/Neu) → Negative wins

# Example 4: Programmatic use with row names
result <- WeightedVote(vote_data, weights)
names(result)  # Returns row names from vote_data
table(result)   # Count votes per category
} # }