An enhancement methodology for predicting transaction amounts for freelancers on freelance platforms: Based on sentiment analysis

Freelancing is a type of labor arrangement in which independent freelancers use their discretionary time to perform ad hoc tasks, usually of a single nature, with the aim of getting paid. This study proposes an augmented method for predicting the amount of freelance transactions using textual sentiment analysis. First, a unique feature labeled Freelancer Sentiment was selected to summarize the positive or negative sentiment orientation of freelancers. Subsequently, Naive Bayes algorithm is applied to process the text data from the freelancers platform to finally develop a Model for Computing Word Sentiment Values. The model helps to accurately calculate the sentiment values associated with emotional words. Finally, the Word Frequency-Inverse Document Frequency (TF-IDF) algorithm is used to construct the Text Sentiment Value Calculation Model, so as to accurately calculate the sentiment values of freelancers. The results of the comparison experiments of the five commonly used prediction models show that the mean squared error (MSE) of the model that includes the freelancer sentiment feature is significantly reduced by 6%-11% compared with the model that does not include the freelancer sentiment feature. This study contains theoretical explorations and practical implications. First, the proposed approach of extracting features from textual data to build predictive models provides a valuable reference for future enhancement of predictive modeling on freelance platforms, especially those that rely on unstructured data. Second, incorporating textual sentiment value features relevant to freelancers can significantly improve the accuracy of predicting transaction amounts. Third, the calculation of word and text sentiment values employs a series of algorithms that target specific features of the freelancer platforms text data. This approach is important for improving the accuracy of feature value calculation.