This paper presents EventAction, a visual analytics system for prescriptive analytics on temporal event sequences. The workflow helps analysts recommend actions to improve outcomes by leveraging historical records. Review Current Record. The analyst loads the current subject's temporal event sequence and reviews it as a timeline table showing event categories over time periods. Find Similar Archived Records. The system computes similarity scores between the current record and each archived record based on event sequence feature distances. The analyst examines the similarity distribution and interactively selects a cohort of similar records by adjusting a similarity threshold, guided by indicators including cohort size, proportion with the desired outcome, and average similarity. Explore Outcomes and Recommendations. The system computes outcome distributions relative to baseline, event-outcome correlations per category, and aggregated temporal activity patterns for the cohort and the desired-outcome subgroup. Coordinated views display outcome probability bars, per-category correlation charts with trend encoding, and an activity summary integrated with the current record's timeline — filterable by overall cohort, desired-outcome subgroup, and distinguishing activities. The analyst identifies which event categories are most associated with the desired outcome and when they typically occur. Plan Specification and Iterative Tuning. The analyst creates an initial action plan by specifying planned events by category and time period, guided by identified recommendation patterns. The system recomputes similarity with the extended record and updates outcome probability estimates. The analyst assesses whether the plan achieves sufficient probability improvement; if not, they refine the plan based on outcome feedback and recommendation patterns. This cycle continues until the analyst is satisfied with the plan's estimated impact. Finalize Guidance. The analyst integrates the data-driven action plan with domain expertise to formulate actionable guidance with associated outcome probability estimation.