workflow SOMFlow template: define-unit -> define-unit -> loop(characterise -> build-model -> define-unit -> characterise -> characterise -> characterise -> characterise -> contextualise -> define-unit -> contextualise -> visualise -> visualise -> abstract -> characterise -> assess -> generate-knowledge (specify) -> define-unit -> define-unit) -> generate-knowledge description: "Assumed composite workflow enabled by SOMFlow for guided exploratory cluster analysis of time-series collections. Analysts iteratively inspect topology-preserving cluster maps, assess quality and metadata relations, adapt analysis settings, partition data into subsets, and reflect on branching analytical provenance." # ============================================================================ # INPUT # ============================================================================ artifact D_time_series : entities origin: given internal structure: sequence embedment: set features: - id: primary_series value structure: vector value type: numeric description: "Numeric sequence representing the primary measurement profile of each data item" - id: metadata_attributes value structure: vector value type: categorical description: "Descriptive categorical attributes associated with each time-series item" description: "Collection of time-series items with associated metadata" artifact S_exploration_configuration : specification origin: given representation form: "preprocessing, clustering, partitioning, layout, and display settings" description: "Initial settings for data transformation, feature weighting, similarity calculation, cluster-map configuration, cell and provenance layouts, metadata overlays, recommendation thresholds, and partitioning strategy" # ============================================================================ # INITIALISE ANALYTICAL FOCUS AND PROVENANCE # ============================================================================ transform T_initialise_analysis_subset : intent: define-unit manner: "initialise analysis collection" input: D_time_series output: D_current_subset actor: machine description: "Create the initial current subset containing the entire time-series collection for exploratory analysis" artifact D_current_subset : entities internal structure: sequence embedment: set description: "Current collection of time-series items selected for cluster-map analysis" transform T_initialise_provenance : intent: define-unit manner: "create analytical provenance structure" input: D_current_subset output: D_analysis_graph, D_analysis_states actor: machine description: "Create a provenance graph and its initial analysis state for recording cluster-map results, partitions, and their relationships" artifact D_analysis_graph : entities internal structure: formation description: "Single analytical provenance graph containing analysis states and links representing subset-selection and partitioning relationships" artifact D_analysis_states : entities internal structure: elementary embedment: relational structure description: "Analysis-state nodes contained in the provenance graph, each representing a recorded cluster-map result and its associated subset" # ============================================================================ # GUIDED EXPLORATORY CLUSTER ANALYSIS # ============================================================================ loop L_guided_cluster_exploration: purpose: "Iteratively analyse cluster-map results, adapt computations, create finer-grained data partitions, and preserve an overview of alternative analytical branches" until: "Relevant structures, metadata relations, data-quality issues, and useful data partitions have been sufficiently explored and documented" body: # ======================================================================== # PREPARE CURRENT DATA SUBSET # ======================================================================== transform T_prepare_series_representation : intent: characterise manner: "preprocess and weight time-series profiles" input: D_current_subset, S_exploration_configuration output: F_series_representation actor: machine description: "Transform, align, and weight time-series profiles according to the current analysis configuration" artifact F_series_representation : feature(D_current_subset) value structure: vector value type: numeric description: "Preprocessed and weighted numeric representation of each time-series item for similarity-based analysis" # ======================================================================== # BUILD AND CHARACTERISE CURRENT CLUSTER MAP # ======================================================================== transform T_build_cluster_map : intent: build-model manner: "fit topology-preserving clustering model" input: F_series_representation, S_exploration_configuration output: M_cluster_map actor: machine description: "Construct a two-dimensional prototype-based clustering model for the current time-series subset" artifact M_cluster_map : model(F_series_representation, S_exploration_configuration) model type: "topology-preserving clustering model" representation form: "two-dimensional grid of prototype profiles" description: "Model representing the current time-series subset by a two-dimensional arrangement of representative profiles" transform T_define_map_cells : intent: define-unit manner: "derive prototype-cell units" input: M_cluster_map output: D_map_cells actor: machine description: "Create cluster-map cells as units representing local regions of the learned profile space" artifact D_map_cells : entities internal structure: elementary embedment: space description: "Cells of the two-dimensional cluster map, each representing a prototype profile and its associated data items" transform T_characterise_map_cells : intent: characterise manner: "profile prototype cells" input: D_map_cells, M_cluster_map, F_series_representation output: F_cell_profile, F_cell_density, F_cell_variation actor: machine description: "Compute representative profiles, membership counts, and within-cell variation for cluster-map cells" artifact F_cell_profile : feature(D_map_cells) value structure: vector value type: numeric description: "Representative time-series profile for each cluster-map cell" artifact F_cell_density : feature(D_map_cells) value structure: atomic value type: numeric description: "Number or relative density of time-series items associated with each cluster-map cell" artifact F_cell_variation : feature(D_map_cells) value structure: atomic value type: numeric description: "Within-cell profile variation indicating local aggregation precision" transform T_assign_items_to_cells : intent: characterise manner: "assign items to prototype cells" input: D_current_subset, F_series_representation, M_cluster_map output: F_item_cell_membership actor: machine description: "Associate each time-series item with its best-matching cluster-map cell" artifact F_item_cell_membership : feature(D_current_subset) value structure: atomic value type: reference description: "Reference to the cluster-map cell associated with each time-series item" transform T_characterise_cell_guidance : intent: characterise manner: "compute topology, quality, and metadata-relation features" input: D_map_cells, D_current_subset, F_cell_profile, F_cell_density, F_cell_variation, F_item_cell_membership output: F_cell_topology, F_cell_quality, F_metadata_association actor: machine description: "Compute cell-neighbourhood relations, separation indicators, and associations between cell membership and metadata attributes" artifact F_cell_topology : feature(D_map_cells) value structure: relational configuration representation form: "neighbourhood relations between map cells" description: "Similarity and neighbourhood relationships among cluster-map cells" artifact F_cell_quality : feature(D_map_cells) value structure: vector value type: numeric description: "Cell-level indicators of separation from neighbouring cells and potential cluster boundaries" artifact F_metadata_association : feature(D_map_cells) value structure: vector value type: numeric description: "Strength of association between cell membership and available metadata attributes" transform T_compute_partition_cues : intent: characterise manner: "rank candidate refinement and partitioning actions" input: D_map_cells, F_cell_topology, F_cell_quality, F_metadata_association output: F_partition_cues actor: machine description: "Derive ranked recommendations for refining imprecise regions, splitting by metadata relations, or partitioning by cell-group structure" artifact F_partition_cues : feature(D_map_cells) value structure: list value type: categorical description: "Ranked candidate actions for refining or partitioning the current data subset" # ======================================================================== # CONTEXTUALISE CELLS AND RECORD CURRENT ANALYSIS STATE # ======================================================================== transform T_layout_map_cells : intent: contextualise manner: "grid-based or topology-driven cell layout" input: D_map_cells, F_cell_topology, S_exploration_configuration output: D_cell_layout_space, A_map_cell_layout actor: machine description: "Position map cells in a two-dimensional display space using either native grid coordinates or a topology-driven layout based on inter-cell relations" artifact D_cell_layout_space : entities internal structure: elementary features: - id: dimensions value structure: atomic value type: numeric description: "Number of display dimensions" description: "Computed two-dimensional display space for the cluster-map overview" artifact A_map_cell_layout : arrangement(D_map_cells) context: D_cell_layout_space principle: "native grid position or topology-driven layout based on cell neighbourhood relations" description: "Placement of map cells in either their default grid positions or a rearranged similarity-oriented layout" transform T_register_analysis_state : intent: define-unit manner: "extend analytical provenance graph" input: D_analysis_graph, D_analysis_states, D_current_subset, M_cluster_map, D_map_cells, F_item_cell_membership output: D_analysis_graph_r, D_analysis_states_r actor: machine description: "Add the current cluster-map analysis state and its relationship to the analysed subset to the provenance graph" artifact D_analysis_graph_r : entities internal structure: formation description: "Updated provenance graph containing the current cluster-map analysis state and its connections to preceding analysis states" artifact D_analysis_states_r : entities internal structure: elementary embedment: relational structure description: "Updated collection of provenance-state nodes, including the current cluster-map analysis state" assign: D_analysis_graph := D_analysis_graph_r D_analysis_states := D_analysis_states_r transform T_layout_analysis_graph : intent: contextualise manner: "force-directed analytical provenance layout" input: D_analysis_graph, D_analysis_states, S_exploration_configuration output: D_analysis_flow_layout_space, A_analysis_state_layout actor: machine description: "Position provenance-state nodes in a two-dimensional flow layout that exposes parent-child analysis relationships and supports comparison of alternative branches" artifact D_analysis_flow_layout_space : entities internal structure: elementary features: - id: dimensions value structure: atomic value type: numeric description: "Number of display dimensions" description: "Computed two-dimensional display space for the analytical provenance view" artifact A_analysis_state_layout : arrangement(D_analysis_states) context: D_analysis_flow_layout_space principle: "force-directed placement based on provenance links" description: "Placement of recorded analysis states to reveal derivation relationships and branching structure" # ======================================================================== # VISUAL ANALYSIS OF CURRENT RESULT AND ANALYTICAL HISTORY # ======================================================================== transform T_visualise_cluster_map : intent: visualise manner: "interactive cluster-map overview" input: D_map_cells, D_current_subset, A_map_cell_layout, F_cell_profile, F_cell_density, F_cell_variation, F_cell_topology, F_cell_quality, F_metadata_association, F_partition_cues, F_item_cell_membership, S_exploration_configuration output: V_cluster_map actor: machine description: "Display representative profiles, cell density, topology, quality overlays, metadata associations, and guidance cues for the current cluster map" artifact V_cluster_map : visualisation(D_map_cells, D_current_subset, A_map_cell_layout, F_cell_profile, F_cell_density, F_cell_variation, F_cell_topology, F_cell_quality, F_metadata_association, F_partition_cues, F_item_cell_membership) layout: "two-dimensional grid or topology-driven cluster-map layout" form: "profile glyphs, cell overlays, borders, metadata marks, recommendation cues, and linked item displays" encoding: "cell position from A_map_cell_layout; profile marks from F_cell_profile; cell size or colour from F_cell_density; variation overlay from F_cell_variation; separation overlay from F_cell_quality; metadata colouring from F_metadata_association; suggested actions from F_partition_cues" description: "Interactive display for inspecting cluster-map structure, local quality, metadata relationships, and candidate refinement actions" transform T_visualise_analysis_provenance : intent: visualise manner: "interactive analytical provenance graph" input: D_analysis_graph, D_analysis_states, A_analysis_state_layout, F_item_cell_membership, F_metadata_association output: V_analysis_flow actor: machine description: "Display the branching history of cluster-map analyses and data partitions to support comparison, navigation, and reflection on prior analytical decisions" artifact V_analysis_flow : visualisation(D_analysis_graph, D_analysis_states, A_analysis_state_layout, F_item_cell_membership, F_metadata_association) layout: "force-directed analytical flow graph" form: "cluster-map nodes, partition links, annotations, and linked highlights" encoding: "state-node position from A_analysis_state_layout; graph nodes represent recorded analysis-state snapshots; links represent subset derivation and partitioning; link prominence represents relative data flow; shared-item highlighting derives from F_item_cell_membership" description: "Interactive provenance view showing the analytical branching structure, preserved analysis snapshots, and relationships among cluster-map results" transform T_interpret_cluster_patterns : intent: abstract manner: "perception-and-interpretation" input: V_cluster_map, V_analysis_flow, F_cell_profile, F_cell_quality, F_metadata_association output: P_cluster_patterns actor: human description: "Identify meaningful profile groups, metadata-related structures, heterogeneous regions, data-quality issues, and promising branches for further exploration" artifact P_cluster_patterns : pattern(D_map_cells, F_cell_profile, F_metadata_association) representation form: "profile-group descriptions and relation annotations" description: "Interpreted cluster-map structures, profile characteristics, metadata relations, and candidate subsets of analytical interest" transform T_annotate_current_result : intent: characterise manner: "manual annotation of cells and items" input: D_current_subset, D_map_cells, V_cluster_map, P_cluster_patterns output: F_manual_annotations, F_cell_annotations actor: human description: "Assign analyst-defined labels to selected cells or time-series items for documenting interpretations, data issues, or analytical relevance" artifact F_manual_annotations : feature(D_current_subset) value structure: list value type: categorical description: "Analyst-defined labels marking time-series items as relevant, uninteresting, erroneous, or otherwise notable" artifact F_cell_annotations : feature(D_map_cells) value structure: list value type: categorical description: "Analyst-defined labels marking cluster-map cells as noisy, relevant, uninteresting, or otherwise notable" transform T_assess_current_analysis : intent: assess manner: "evaluate cluster-map adequacy and exploration value" input: V_cluster_map, V_analysis_flow, P_cluster_patterns, F_cell_quality, F_partition_cues, F_manual_annotations, F_cell_annotations output: K_analysis_assessment actor: human description: "Assess whether the current result adequately reveals relevant structures and whether further computation adaptation, data partitioning, or provenance exploration is needed" artifact K_analysis_assessment : knowledge(P_cluster_patterns, D_map_cells) representation form: "quality judgment" description: "Assessment of cluster-map quality, metadata relevance, usefulness of guidance, and need for further exploration" if K_analysis_assessment indicates further exploration needed: then: transform T_specify_next_action : intent: generate-knowledge manner: "select refinement and partitioning strategy" input: K_analysis_assessment, P_cluster_patterns, V_cluster_map, V_analysis_flow, S_exploration_configuration output: S_exploration_configuration_r actor: human description: "Specify updated preprocessing, feature weighting, cluster-map configuration, layout, metadata focus, or partitioning action for the next analysis branch" artifact S_exploration_configuration_r : specification representation form: "preprocessing, clustering, partitioning, layout, and display settings" description: "Updated settings defining the next computation, layout, partitioning, and visual exploration action" transform T_define_next_subset : intent: define-unit manner: "select or partition time-series subset" input: D_current_subset, D_map_cells, F_item_cell_membership, F_cell_quality, F_metadata_association, F_partition_cues, F_manual_annotations, F_cell_annotations, S_exploration_configuration_r output: D_next_subset actor: hybrid description: "Create the next analysis subset by selecting map cells, cluster regions, metadata-defined groups, labelled items, or items requiring finer-grained refinement" artifact D_next_subset : entities internal structure: sequence embedment: set description: "Selected or partitioned subset of time-series items for the next cluster-map analysis" transform T_extend_partition_provenance : intent: define-unit manner: "record subset derivation in provenance graph" input: D_analysis_graph, D_analysis_states, D_current_subset, D_next_subset, P_cluster_patterns, F_manual_annotations, F_cell_annotations, S_exploration_configuration_r output: D_analysis_graph_next, D_analysis_states_next actor: machine description: "Add the selected partitioning action and its parent-child subset relationship to the provenance graph" artifact D_analysis_graph_next : entities internal structure: formation description: "Updated provenance graph recording the selected next branch and its derivation from the current analysis state" artifact D_analysis_states_next : entities internal structure: elementary embedment: relational structure description: "Updated provenance-state collection recording the newly created analysis branch" assign: S_exploration_configuration := S_exploration_configuration_r D_current_subset := D_next_subset D_analysis_graph := D_analysis_graph_next D_analysis_states := D_analysis_states_next else: exit loop L_guided_cluster_exploration end loop L_guided_cluster_exploration # ============================================================================ # KNOWLEDGE SYNTHESIS # ============================================================================ transform T_synthesise_cluster_findings : intent: generate-knowledge manner: "formulate statements about profile groups and metadata relations" input: P_cluster_patterns, K_analysis_assessment, V_cluster_map, V_analysis_flow, F_manual_annotations, F_cell_annotations output: K_cluster_findings actor: human description: "Document findings about meaningful time-series profile groups, metadata relations, data-quality issues, and the analytical decisions represented in the provenance graph" artifact K_cluster_findings : knowledge(P_cluster_patterns, D_analysis_graph) representation form: "statements, explanations, and annotated analytical provenance" description: "Documented findings about discovered profile structures, relevant metadata relations, selected data partitions, and the reasoning underlying the exploratory analysis"