CSIS Evidence-to-Action Decision Support Prototype

Evidence-to-action recommendations generated from the active dataset and its refreshed analytical outputs, subject to LGU validation.

Decision Rules Summary

Select a row to review its recommendation, supporting evidence, and traceability path.

7 service-area rules
Rank Service Area Service Stage Priority Concern Signal Priority
1 Economic and Investment Promotion 5 cluster segments Awareness Economic and Investment Promotion awareness gap Awareness Gap is the strongest current signal for Organization,accreditation and training of tourism related concessions. (76.7%). 76.7% High
2 Public Works and Infrastructure Home-secure citizens with lower service awareness Awareness Public Works and Infrastructure awareness gap Awareness Gap is the strongest current signal for Information and reading center (70.7%). 70.7% High
3 Support to Education Younger or single citizens facing service-access gaps Availment Support to Education availment gap Availment Gap is the strongest current signal for Alternative Learning System and/or other Special Education Programs (64.8%). 64.8% High
4 Social Welfare Services Home-secure citizens requesting further LGU action Need for Action Social Welfare Services need for action Need for Action is the strongest current signal for Older Persons / Senior Citizens Program (60.7%). 60.7% High
5 Environmental Management Younger or single citizens requesting further LGU action Need for Action Environmental Management need for action Need for Action is the strongest current signal for Solid Waste Management (54.7%). 54.7% High
6 Health Services Younger or single citizens requesting further LGU action Need for Action Health Services need for action Need for Action is the strongest current signal for Free Basic Medicine or Low-Cost Medicine Program (52.4%). 52.4% High
7 Governance and Response Younger or single citizens requesting further LGU action Need for Action Governance and Response need for action Need for Action is the strongest current signal for Traffic Management (51.5%). 51.5% High
Selected Decision
Health Services Cluster 2 Need for Action

Prioritize barangay-level health follow-up for underserved Cluster 2 citizens.

Use targeted outreach and referral tracking where citizens report unmet health-service needs but low actual availment.

42% Strongest current service-gap signal
0.78 How reliably the model separates response groups (0 = weak, 1 = strong)
3 Linked analytical sources: clustering, Decision Tree, and Apriori
Observed concern D11 responses show the strongest need-for-action signal.
Group to review first The selected cluster has the strongest related concern signal.
Proposed next step Validate the finding with the responsible service office and LGU records.
Decision status: AI-assisted recommendation pending human review. The system organizes analytical evidence and proposes a planning response; it does not make or approve policy decisions. Confirm local context, feasibility, and responsible-office approval before implementation.

Traceability

Follow the selected recommendation from CSIS survey evidence to a proposed LGU action, or reverse the flow for an audit.

1 Survey evidence D9 and D11 service-stage responses.
2 Model evidence Cluster summary, Decision Tree metrics, and Apriori pattern.
3 Decision rule High need for action with low availment.
4 Proposed LGU action Barangay-level health follow-up for Cluster 2.
Human validation required. Confirm the proposed action using LGU records, service-office evidence, feasibility review, field context, and stakeholder approval.
Detailed Evidence-to-Action Review Health Services · 5 indicator records Sankey and evidence audit
Analytics -> Harmonization -> Human Review -> Action

Evidence-to-Action Review

Cross-stage, clustering, Decision Tree, and Apriori results are translated into proposed LGU actions. Recommendations remain subject to local validation, feasibility review, and approval.

Health Services: Priorities are generated from the active dataset's largest indicator-level service gaps. Segment labels identify the portal group with the strongest related concern signal.
Every decision-support record includes: rank, cross-stage result, clustering result, Decision Tree evidence, repeated service-stage pattern, recommended LGU action, priority, responsible office, evidence audit, and monitoring indicator.
Recurring Decision Tree predictors — exploratory consistency signal: MCA Dim2 · 3/5 models (60%) MCA DimMagnitude · 3/5 models (60%) House Ownership · 3/5 models (60%) MCA Dim1 · 2/5 models (40%)
Repeated selection can identify factors for LGU investigation, but it does not overcome weak model discrimination, establish causality, or support individual targeting. Frequencies use available indicator models with at least one non-zero feature.
Cross-Service Benchmark

Predictor Coherence

Checks whether recurring Health Services predictors also recur across the wider CSIS classification workflow.

Open cross-service benchmark
Predictor Health Services recurrence Cross-service coverage Cross-Service Coherence
MCA Dim2 3/5 models (60%) 143 models · 7 service areas · 4 stages System-wide coherent
The predictor recurs across at least three service areas and two service stages.
MCA DimMagnitude 3/5 models (60%) 118 models · 7 service areas · 4 stages System-wide coherent
The predictor recurs across at least three service areas and two service stages.
House Ownership 3/5 models (60%) 66 models · 7 service areas · 4 stages System-wide coherent
The predictor recurs across at least three service areas and two service stages.
MCA Dim1 2/5 models (40%) 133 models · 7 service areas · 4 stages System-wide coherent
The predictor recurs across at least three service areas and two service stages.
System-wide coherent means recurrence across at least three service areas and two service stages. Feature importance is unsigned, so this comparison checks recurrence and coverage—not relationship direction, causality, or suitability for individual targeting.
Dynamic indicator flow

Free Basic Medicine or Low-Cost Medicine Program

Hover over the Sankey nodes to audit how the four analytical evidence sources converge into an integrated interpretation, priority, proposed action, and monitoring measure.
Rank Cross-Stage Result Clustering Result Decision Tree Evidence Repeated Service-Stage Pattern Recommended LGU Action Priority Implementation and Monitoring
1 Free Basic Medicine or Low-Cost Medicine Program
Need for Action (52.4%)
Citizens who used the service still indicate a need for government action.
Awareness 87.7%; availment 75.7%; satisfaction 91.1%; need for action 52.4%.
 Young single student citizens (Segment B) (60.6%)
Young single student citizens (Segment B) has the highest concern signal at 60.6% for this indicator.
 Model quality
Needs more predictors: F1 37.3%, ROC AUC 48.0%. The model has difficulty separating respondents based on ROC AUC, recall, or F1.
Leading feature (exploratory)
MCA Dim1 (42.3% importance)
Treat as a validation cue, not a reliable explanation or causal factor.
 Respondents who were aware of the healthservices, and availed the healthservices had a 98% likelihood of also were satisfied with the healthservices (Lift: 1.02). Support 59.7%, confidence 98.0%, lift 1.02. Prioritize follow-up validation with the service office and compare findings with LGU service records. Critical Municipal/City Health Office
Monitor: Reduction in need-for-action percentage
2 Basic dental/oral hygiene
Need for Action (49.4%)
Citizens who used the service still indicate a need for government action.
Awareness 77.2%; availment 45.6%; satisfaction 95.5%; need for action 49.4%.
 Working male homeowners (Segment A) (53.6%)
Working male homeowners (Segment A) has the highest concern signal at 53.6% for this indicator.
 Model quality
Needs more predictors: F1 50.2%, ROC AUC 49.4%. The model has difficulty separating respondents based on ROC AUC, recall, or F1.
Leading feature (exploratory)
MCA Dim1 (85.1% importance)
Treat as a validation cue, not a reliable explanation or causal factor.
 Respondents who were aware of the healthservices, and availed the healthservices had a 98% likelihood of also were satisfied with the healthservices (Lift: 1.02). Support 59.7%, confidence 98.0%, lift 1.02. Prioritize follow-up validation with the service office and compare findings with LGU service records. High Municipal/City Health Office
Monitor: Reduction in need-for-action percentage
3 Free General Consultations/Access to secondary and/or tertiary health care
Need for Action (47.7%)
Citizens who used the service still indicate a need for government action.
Awareness 87.7%; availment 71.5%; satisfaction 95.5%; need for action 47.7%.
 Young single student citizens (Segment B) (56.0%)
Young single student citizens (Segment B) has the highest concern signal at 56.0% for this indicator.
 Model quality
Needs more predictors: F1 49.8%, ROC AUC 50.5%. The model has difficulty separating respondents based on ROC AUC, recall, or F1.
Leading feature (exploratory)
MCA Dim2 (71.3% importance)
Treat as a validation cue, not a reliable explanation or causal factor.
 Respondents who were aware of the healthservices, and availed the healthservices had a 98% likelihood of also were satisfied with the healthservices (Lift: 1.02). Support 59.7%, confidence 98.0%, lift 1.02. Prioritize follow-up validation with the service office and compare findings with LGU service records. High Municipal/City Health Office
Monitor: Reduction in need-for-action percentage
4 Family Planning/ Reproductive Health Distribution of reproductive health supplies, information dissemination and other services
Need for Action (46.0%)
Citizens who used the service still indicate a need for government action.
Awareness 86.2%; availment 51.9%; satisfaction 97.5%; need for action 46.0%.
 Working male homeowners (Segment A) (48.3%)
Working male homeowners (Segment A) has the highest concern signal at 48.3% for this indicator.
 Model quality
Limited signal: F1 62.3%, ROC AUC 52.0%. The model can be reviewed as an exploratory clue, but F1, ROC AUC, or baseline gain is not strong enough for confident prediction.
Leading feature (exploratory)
MCA Dim2 (49.7% importance)
Treat as a validation cue, not a reliable explanation or causal factor.
 Respondents who were aware of the healthservices, and availed the healthservices had a 98% likelihood of also were satisfied with the healthservices (Lift: 1.02). Support 59.7%, confidence 98.0%, lift 1.02. Prioritize follow-up validation with the service office and compare findings with LGU service records. High Municipal/City Health Office
Monitor: Reduction in need-for-action percentage
5 Vaccination for infants/children
Need for Action (45.1%)
Citizens who used the service still indicate a need for government action.
Awareness 95.2%; availment 66.2%; satisfaction 98.7%; need for action 45.1%.
 Young single student citizens (Segment B) (50.0%)
Young single student citizens (Segment B) has the highest concern signal at 50.0% for this indicator.
 Model quality
Needs more predictors: F1 57.2%, ROC AUC 47.1%. The model has difficulty separating respondents based on ROC AUC, recall, or F1.
Leading feature (exploratory)
House Ownership (40.0% importance)
Treat as a validation cue, not a reliable explanation or causal factor.
 Respondents who were aware of the healthservices, and availed the healthservices had a 98% likelihood of also were satisfied with the healthservices (Lift: 1.02). Support 59.7%, confidence 98.0%, lift 1.02. Prioritize follow-up validation with the service office and compare findings with LGU service records. High Municipal/City Health Office
Monitor: Reduction in need-for-action percentage
Priority method: Critical for satisfaction or need-for-action gaps of at least 50%; High for gaps of at least 40%; Medium for gaps of at least 20%; otherwise Low. Recommendations are evidence-based prompts—not automatic policy decisions.