Overview
The Activity Frequency filter removes activities that appear too rarely or too frequently across your process cases. Unlike case-level filters that remove entire cases, this event-level filter analyzes how often each activity appears across all cases and removes individual events for activities that fall outside your specified frequency range. This helps you focus on activities that matter most by removing noise from rare exceptions or filtering out overly common activities that don't provide analytical value.
The filter calculates the percentage of cases that contain each activity, then removes all events for activities whose frequency falls outside the minimum and maximum thresholds you specify. This is particularly useful for simplifying process maps, focusing analysis on meaningful activities, and removing data quality issues caused by rare or unusual events.
Common Uses
- Process Simplification: Remove very rare activities that clutter process maps and make analysis difficult.
- Noise Reduction: Filter out exceptional activities that appear in less than 5% of cases to focus on standard process flows.
- Core Process Analysis: Analyze only the most common activities by filtering to activities that appear in 80% or more of cases.
- Middle-Range Focus: Examine activities with moderate frequency (e.g., 20-80%) to identify optional steps or exceptions.
- Data Quality Assessment: Identify and remove activities with unusual frequency patterns that may indicate data quality issues.
- Process Map Clarity: Create cleaner process maps by removing both very rare and very common activities that don't add analytical value.
Settings
Minimum Percent: The minimum percentage of cases (0.0 to 1.0) an activity must appear in to be included. Activities appearing in fewer cases will be filtered out. For example, 0.2 means the activity must appear in at least 20% of cases.
Maximum Percent: The maximum percentage of cases (0.0 to 1.0) an activity can appear in to be included. Activities appearing in more cases will be filtered out. For example, 0.8 means the activity must appear in no more than 80% of cases.
Note: The filter uses inclusive range checking, so activities with frequencies exactly equal to the minimum or maximum percentages will be included. Both minimum and maximum must be values between 0.0 (0%) and 1.0 (100%).
Examples
Example 1: Removing Rare Exception Activities
Scenario: Your purchase order process contains many rare exception activities that clutter your process map. You want to focus on standard activities that appear in at least 10% of cases.
Settings:
- Minimum Percent: 0.1
- Maximum Percent: 1.0
Result: All activities that appear in fewer than 10% of cases are removed. For example, if "Emergency Approval" only appears in 5% of cases, all events with that activity are filtered out.
Insights: This creates a cleaner view of your standard process flow by removing rare exceptions like emergency procedures, special escalations, or unusual corrections. You can then analyze these exceptional cases separately if needed.
Example 2: Focusing on Core Process Activities
Scenario: You want to analyze only the core activities that occur in nearly all cases, filtering out optional or conditional steps.
Settings:
- Minimum Percent: 0.8
- Maximum Percent: 1.0
Result: Only activities that appear in 80% or more of cases are retained. Activities like "Create Order" (100% of cases), "Approve Order" (95% of cases), and "Ship Order" (85% of cases) are kept, while optional activities like "Apply Discount" (40% of cases) are removed.
Insights: This reveals your mandatory process steps and standard path, helping you understand the core workflow that most cases follow. Deviations from this core can be analyzed separately.
Example 3: Analyzing Mid-Frequency Activities
Scenario: You want to focus on activities that appear in a moderate number of cases (20-80%) to understand optional process steps and common variations.
Settings:
- Minimum Percent: 0.2
- Maximum Percent: 0.8
Result: Very rare activities (under 20%) and very common activities (over 80%) are removed, leaving only mid-frequency activities.
Insights: This helps identify:
- Optional process steps that are frequently but not universally used
- Common process variations that occur in a substantial minority of cases
- Activities that may be candidates for standardization or removal
- Process branches that serve specific customer segments or product types
Example 4: Removing Ubiquitous Activities
Scenario: Your process has some administrative activities that appear in nearly every case but don't provide analytical insights. You want to remove activities that appear in more than 95% of cases.
Settings:
- Minimum Percent: 0.0
- Maximum Percent: 0.95
Result: Activities that appear in more than 95% of cases are removed. For example, if "System Log Entry" appears in 99% of cases, all those events are filtered out.
Insights: This removes activities that occur so frequently they don't help differentiate between cases or process paths. It helps focus on activities that actually indicate process variations or decisions.
Example 5: Finding Activities at Specific Frequency
Scenario: You want to analyze only activities that appear in exactly 50% of cases (plus or minus a small margin) to understand process branching points.
Settings:
- Minimum Percent: 0.45
- Maximum Percent: 0.55
Result: Only activities that appear in 45-55% of cases are retained. These often represent decision points where the process splits into two roughly equal paths.
Insights: These activities typically indicate:
- Binary process decisions (approved/rejected, domestic/international)
- Optional features chosen by approximately half of customers
- Seasonal variations that affect half the year's cases
- Process changes that were implemented mid-period
Example 6: Comprehensive Noise Reduction
Scenario: You want to remove both very rare exceptions (under 5%) and very common administrative activities (over 90%) to focus on meaningful process activities.
Settings:
- Minimum Percent: 0.05
- Maximum Percent: 0.9
Result: The filter removes rare exception activities and ubiquitous administrative activities, leaving activities that appear in 5-90% of cases.
Insights: This creates a balanced view that:
- Excludes rare data quality issues and exceptional cases
- Removes administrative overhead activities
- Retains all meaningful business process activities
- Provides a clear picture for process optimization analysis
Output
The filter returns a new dataset containing only events for activities whose frequency falls within the specified range. Cases may have fewer events after filtering, but no entire cases are removed unless all their activities fall outside the frequency range.
If you set Minimum Percent to 0.0 and Maximum Percent to 1.0, no filtering occurs and all activities are retained.
The output preserves all event attributes and timestamps for the retained events, maintaining temporal and contextual information for the filtered process data.
Technical Notes
- Filter Type: Event-level filter (removes individual events, which affects cases)
- Frequency Calculation: Counts the number of cases containing each activity, not the number of times the activity occurs
- Performance: Analyzes all activities and cases, then filters events based on frequency calculation
- Empty Cases: Cases may become empty if all their activities are filtered out
- Percentage Format: Uses decimal format (0.0 to 1.0) rather than percentage format (0 to 100)
This documentation is part of the mindzieStudio process mining platform.