Methodology for Risk Assessment and Analytical Indices

1. General Principles

The analytical system is based on multi-layer event-driven data processing and is designed to identify, quantify, and monitor risks within a dynamic information environment. The methodology integrates elements of statistical analysis, applied linguistics, signal theory, and probabilistic forecasting.

From an architectural perspective, the system is divided into two logically independent yet coordinated levels:

1. Risk Index — an aggregated macro-indicator reflecting the current level of threats and their dynamics.
2. KASVI MODEL — an applied operational analysis model used during data parsing and frontend-level visualization.

2. Data Sources and Characteristics

All calculations rely exclusively on open and publicly available data streams, processed in near-real time:

• textual content (news materials, social media, public channels),
• temporal and contextual metadata,
• structural dissemination features,
• semantic and thematic content attributes.

The system does not process personal data and does not perform individual profiling.

3. The Risk Index Model

3.1 Purpose

The Risk Index is intended for integrated assessment of potential threat levels at a given point in time and serves as a tool for strategic monitoring and comparative analysis.

3.2 Dimensions

The index is computed using a set of independent analytical dimensions, each capturing a distinct aspect of risk:

• information flow intensity — frequency and velocity of relevant signals,
• content-related threat — presence of indicators of incidents, escalation, violence, or destabilization,
• reliability — degree of corroboration and source quality,
• structural activity — synchronicity, repetition, and coordination patterns,
• contextual relevance — spatial-temporal linkage and continuity with preceding events,
• predictive component — probabilistic assessment of short-term escalation.

Each dimension is constructed autonomously and does not depend on the final index value.

3.3 Nonlinear Aggregation

The final Risk Index is produced using nonlinear aggregation. This approach avoids simple averaging and enables the system to:

• amplify high-risk signals,
• remain sensitive to rare but critical events,
• prevent dilution of severe signals by large volumes of neutral content.

The specific aggregation functions, amplification parameters, and thresholds are part of the protected methodology.

3.4 Temporal Stability

The index is calculated across multiple temporal windows (short-term, mid-term, and baseline), ensuring responsiveness while maintaining resistance to noise.

4. KASVI MODEL (Frontend Analysis)

4.1 Role and Distinction from the Risk Index

The KASVI MODEL is an applied analytical framework designed for real-time interpretation of data during parsing and frontend presentation.

Unlike the Risk Index, the KASVI MODEL does not function as a macro-aggregator. It operates at the level of individual streams, topics, temporal slices, and structural features of content.

4.2 Functional Components

The KASVI MODEL performs the following core functions:

• automated semantic and thematic annotation,
• classification of signal and event types,
• assessment of activity levels and anomaly detection within data streams,
• computation of auxiliary sub-metrics for visualization,
• generation of explainable indicators without exposing internal decision rules.

4.3 Integration with the Core System

Outputs generated by the KASVI MODEL are:

• used as input signals for deeper server-side analytics,
• employed for visual representation of risk dynamics,
• applied as tools for real-time interpretation and preliminary validation of data.

The KASVI MODEL does not contain full computational logic of the Risk Index and cannot be used to reconstruct protected system parameters.

5. Explainability and Interpretation

Both models adhere to the principle of controlled explainability:

• users are provided with semantic drivers behind score changes,
• dominant topics and activity types are highlighted,
• temporal dynamics and relative factor contributions are shown.

Precise mathematical formulations, weightings, and heuristics remain undisclosed.

6. Protection Against Distortion and Manipulation

The methodology incorporates safeguards against artificial inflation or manipulation of indicators:

• clustering of duplicate and repeated publications,
• limitation of influence from homogeneous sources,
• detection of synchronized template-driven activity,
• independent assessment of reliability and corroboration.

7. Disclosure Limitations

To protect intellectual property and system robustness, the following elements are intentionally not disclosed:

• formulas, coefficients, and threshold values,
• dictionaries, signatures, and coordination-detection heuristics,
• source lists and ranking logic,
• training datasets and annotation methodologies.

8. Scientific and Practical Applicability

The methodology is intended for research and analytical use, supporting the monitoring of information risks and the study of threat dynamics. The system is not designed for automated decision-making in legal, administrative, or operational contexts and serves solely as an analytical support instrument.