Abstract: An online concierge system receives a delivery order containing a list of items, generates a suggested picking sequence for picking the delivery order in a warehouse, and transmits the suggested picking sequence to a mobile device of the shopper. Generating the suggested sequence includes applying a trained item sequence model to the delivery order. Training the item sequence model includes accessing data about a set of historical orders, determining a pairwise distance between each pair of aisles in the warehouse based on the data about the set of historical orders, and generating a distance graph based on the pairwise distance between each pair of aisles in the warehouse. The plurality of nodes represent a plurality of aisles in the warehouse, and the plurality of edges represent pairwise distances between pairs of aisles.

Abstract: An online concierge system generates a suggested picking sequence to reduce the amount of time for a shopper to fulfill an online order of items from a warehouse. The online concierge system determines an average amount of time to sequentially pick items between different aisle pairs for a warehouse based on timestamps from item fulfillment in historical orders. The system generates a distance graph including aisle nodes connected by edges representing the pairwise distance between aisles. The system solves a traveling salesperson problem to generate a ranked order of aisle nodes for each of the historical orders. The system generates a ranked global sequence of aisle nodes based on the plurality of ranked orders of aisle nodes. The system applies the ranked global sequence to new delivery orders to generate the suggested picking sequence for a shopper.

Abstract: An online concierge system generates a suggested picking sequence to reduce the amount of time for a shopper to fulfill an online order of items from a warehouse. The online concierge system determines an average amount of time to sequentially pick items between different aisle pairs for a warehouse based on timestamps from item fulfillment in historical orders. The system generates a distance graph including aisle nodes connected by edges representing the pairwise distance between aisles. The system solves a traveling salesperson problem to generate a ranked order of aisle nodes for each of the historical orders. The system generates a ranked global sequence of aisle nodes based on the plurality of ranked orders of aisle nodes. The system applies the ranked global sequence to new delivery orders to generate the suggested picking sequence for a shopper.

Abstract: Systems and methods for enhanced detection of fraudulent electronic transactions are disclosed. In one embodiment, a system uses the ongoing stream of transactions to construct and maintain a dynamically evolving merchant relationship graph. When a proposed transaction is submitted to the system, the system computes a predicted likelihood that the given account would make a transaction with these characteristics with the given merchant. The graph is used to compute transitive relatedness between merchants which may be indirectly associated with one another, as well as to compute aggregate relatedness, when there are multiple avenues of relationship between two merchants.

Abstract: Systems and methods for generating real-time, personalized recommendations are disclosed. In one embodiment, a method operates upon an electronic data collection organized as a network of vertices and edge connections between the vertices. The method provides the recommendations includes iteratively traversing across edges that satisfy search criteria to a new set of vertices and filtering each new set of vertices to satisfy the search criteria. At the conclusion of the traversing and filtering, a final set of vertices represents the recommended entities. In some embodiments, a control vector describes a sequence of relationships between a requester and the items to be recommended. The method can assign scores to candidate recommendations and select the recommendations having the highest scores. Advantageously, the method provides flexibility and rapid execution of recommendation queries without the need to precompute intermediate results.

Abstract: Embodiments disclosed herein provide systems, methods, and software for determining the shortest distance between vertices in relatively large graphs. In an embodiment a method includes identifying an original graph comprising a plurality of nodes and a plurality of edges, identifying a plurality of hub nodes from within the original graph, creating a hub matrix comprising the plurality of hub nodes and the shortest distance between the plurality of hub nodes, and determining a shortest distance between the at least two nodes using the original graph and/or the hub matrix.

Abstract: Methods and systems for distributed computation of graph data permit edge collection and vertex collection, each to be partitioned among a plurality of computational units. In one embodiment, the methods employ a two-phase computational cycle, which is repeated until the computation is complete. In a first phase, processing units process each active edge and vertex by doing the following: reading their current attribute values, executing programmed computational functions, updating edge attributes and sending data messages to vertices. In a second phase, each vertex update processor processes each of its active vertices by doing the following: reading its current attribute values and received data messages, executing a programmed computational function, and updating the vertex's attribute values.

Abstract: Systems and methods for generating real-time, personalized recommendations are disclosed. In one embodiment, a method operates upon an electronic data collection organized as a network of vertices and edge connections between the vertices. The method provides the recommendations includes iteratively traversing across edges that satisfy search criteria to a new set of vertices and filtering each new set of vertices to satisfy the search criteria. At the conclusion of the traversing and filtering, a final set of vertices represents the recommended entities. In some embodiments, a control vector describes a sequence of relationships between a requester and the items to be recommended. The method can assign scores to candidate recommendations and select the recommendations having the highest scores. Advantageously, the method provides flexibility and rapid execution of recommendation queries without the need to precompute intermediate results.

Abstract: Methods and systems for real-time graphical search for airline flight itineraries that satisfy predetermined criteria (e.g., place and time) using a distributed graph processing system are disclosed. The advantages of the graphical method include: computational work is easily split across multiple processors for parallel processing; the resulting speed is appropriate for real-time personalized search; the method naturally supports multi-segment routes up to any user-specified maximum; the method easily handles constraints or freedoms on connections between flights, such as connection time or transferring to another airport in the same metropolis; and the method is efficient due to focusing only on viable flight segments.

Abstract: Methods and systems for distributed computation of graph data permit edge collection and vertex collection, each to be partitioned among a plurality of computational units. In one embodiment, the methods employ a two-phase computational cycle, which is repeated until the computation is complete. In a first phase, processing units process each active edge and vertex by doing the following: reading their current attribute values, executing programmed computational functions, updating edge attributes and sending data messages to vertices. In a second phase, each vertex update processor processes each of its active vertices by doing the following: reading its current attribute values and received data messages, executing a programmed computational function, and updating the vertex's attribute values.

Abstract: Systems and methods for enhanced detection of fraudulent electronic transactions are disclosed. In one embodiment, a system uses the ongoing stream of transactions to construct and maintain a dynamically evolving merchant relationship graph. When a proposed transaction is submitted to the system, the system computes a predicted likelihood that the given account would make a transaction with these characteristics with the given merchant. The graph is used to compute transitive relatedness between merchants which may be indirectly associated with one another, as well as to compute aggregate relatedness, when there are multiple avenues of relationship between two merchants.

Abstract: Embodiments disclosed herein provide systems and methods for scaling reachability computations on relatively large graphs. In an embodiment, a method provides for scaling reachability computations on relatively large graphs, the method comprising, identifying an initial graph comprising a plurality of vertices and a plurality of edges, processing at least a portion of the plurality of vertices and at least a portion of the plurality of edges to generate a plurality of reachability indices for the at least a portion of the plurality of vertices, and generating a backbone graph comprising a scaled-down version of the initial graph, based at least in part on at least one of the plurality of reachability indices.

Abstract: Embodiments disclosed herein provide systems and methods for a unified analytics environment. In a particular embodiment, a method provides, handling a plurality of relational functions within a relational analytics environment. The method further provides, while handling the plurality of relational functions, encountering at least one graph function that comprises a query intended for a graph analytics environment. The method further provides, in response to encountering the at least one graph function, communicating with the graph analytics environment to handle the at least one graph function.

Abstract: Embodiments disclosed herein provide systems and methods for scaling reachability computations on relatively large graphs. In an embodiment, a method provides for scaling reachability computations on relatively large graphs, the method comprising, identifying an initial graph comprising a plurality of vertices and a plurality of edges, processing at least a portion of the plurality of vertices and at least a portion of the plurality of edges to generate a plurality of reachability indices for the at least a portion of the plurality of vertices, and generating a backbone graph comprising a scaled-down version of the initial graph, based at least in part on at least one of the plurality of reachability indices.

Abstract: Embodiments disclosed herein provide systems, methods, and software for determining the shortest distance between vertices in relatively large graphs. In an embodiment a method includes identifying an original graph comprising a plurality of nodes and a plurality of edges, identifying a plurality of hub nodes from within the original graph, creating a hub matrix comprising the plurality of hub nodes and the shortest distance between the plurality of hub nodes, and determining a shortest distance between the at least two nodes using the original graph and/or the hub matrix.