Fig 1.
Satisfaction with life prediction model.
We use Elastic Net regression to select informative features among sentiment, LIWC and LDA generated topics for the random forest model. The model is trained to fit the self-reported SWL score.
Table 1.
Correlation matrix of affect features and self-reported SWL.
Table 2.
Correlations between the prediction performance of the random forest models using different features and self-reported SWL.
Table 3.
Correlations between the random forest predicted and self-reported CES-D.
Fig 2.
We compare the Facebook activity sentiment scores (FB activities (z-scores)) with the activity sentiment scores in the experience sampling study [11] (experience sampling (z-scores)). Since some activities were not included in the experience sampling study, the corresponding columns are empty.
Table 4.
Facebook activity sentiment.