Enhancing Social Media Impact: Leveraging User Demographics for Text Tone Preference Analysis

Abstract

This study investigates the correlation between user demographics and text tone preferences for social media content aimed at addressing food insecurity among Hispanic households in the United States. Using k-means and hierarchical clustering machine learning models, we analyze preliminary data collected from a research group comprising mentors and students who tested the survey framework. By developing various text tone connotations, we aim to uncover patterns in post preferences among the Hispanic community. Our findings will assist food pantries in crafting more engaging and effective social media content.

Dependencies

Keywords

Food Insecurity, Text tone preferences, K-Means, Hierarchical Clustering, Unsupervised Learning

Data Preprocessing Pipeline

age	gender	ethnicity	race	education	marital_status	income	employment	language	disability	states	sample_1	sample_2	sample_3	sample_4	sample_5	sample_6	sample_7	sample_8
45-54	female	non hispanic	native american	High School	na	$25,000 - $49,999	Employed Part time	both	i do not have a disability	indiana	Persuasive	Simpler	Empathetic	Persuasive	Original	Original	Persuasive	Original
18-24	male	hispanic	white	High School	single	Less than $25,000	Employed Part time	english	i do not have a disability	illinois	Original	Simpler	Empathetic	Simpler	Simpler	Original	Original	Persuasive
25-34	female	non hispanic	multiracial	Associate	single	Less than $25,000	Student	english	i do not have a disability	new York	Original	Original	Simpler	Simpler	Empathetic	Empathetic	Empathetic	Simpler

Figure 1.1 Initial dataset

Combining Post Preferences and Attribute Selection

To simplify data analysis and model training, the melt function was utilized to combine individual post choices from multiple columns ('sample_1' to 'sample_8') into a single 'choice' column. This effectively reduced the dataset's dimensionality while preserving crucial demographic information such as age, gender, ethnicity, education, income, employment status, and disability.Each row in the dataset represents an individual submission, with the 'choice' column indicating the preferred post option. This implementation facilitates the examination of individual user preferences while considering demographic characteristics.

age	gender	ethnicity	education	income	employment	disability	choice
45-54	female	non hispanic	High School	$25,000 - $49,999	Employed Part time	i do not have a disability	Persuasive
18-24	male	hispanic	High School	Less than $25,000	Employed Part time	i do not have a disability	Original
25-34	female	non hispanic	Associate	Less than $25,000	Student	i do not have a disability	Original

Encoding Categories

Dataset 1: Ethnicity numerical encoding

Age Category	Encoded Value (Age)	Income Category	Encoded Value (Income)	Disability Category	Encoded Value (Disability)	Ethnicity Category	Encoded Value (Ethnicity)
18-24	0	Less than 25000	0	No	0	Hispanic	1
25-34	1	25000 - 49999	1	Yes	1	Non-Hispanic	0
35-44	2	50000 - 74999	2	Prefer not to say	-1	Prefer not to say	-1
45-54	3	75000 - 99999	3	-	-	-	-
55-64	4	100000 - 149999	4	-	-	-	-
65 and above	5	150000 or more	5	-	-	-	-
Prefer not to say	-1	Prefer not to say	-1	-	-	-	-

age	ethnicity	income	disability	gender_female	gender_male	gender_non binary	education_Associate	education_Bachelor	education_Doctorate	...	employment_Employed Full time	employment_Employed Part time	employment_Retired	employment_Self employed	employment_Student	employment_Unemployed	choice_Empathetic	choice_Original	choice_Persuasive	choice_Simplier
3	0	1	0	True	False	False	False	False	False	...	False	True	False	False	False	False	False	True	False	False
0	1	0	0	False	True	False	False	False	False	...	False	True	False	False	False	False	False	True	False	False
1	0	0	0	True	False	False	True	False	False	...	False	False	False	False	True	False	False	True	False	False

Dataset 2: label encoding

age	income	gender_female	gender_male	gender_non binary	ethnicity_hispanic	ethnicity_non hispanic	education_Associate	education_Bachelor	education_Doctorate	...	employment_Retired	employment_Self employed	employment_Student	employment_Unemployed	disability_i do not have a disability	disability_undisclosed	choice_Empathetic	choice_Original	choice_Persuasive	choice_Simplier
3	1	True	False	False	False	True	False	False	False	...	False	False	False	False	True	False	False	False	True	False
0	0	False	True	False	True	False	False	False	False	...	False	False	False	False	True	False	False	True	False	False

PCA Implementation