TW201938993A - Operation path navigation - Google Patents
Operation path navigation Download PDFInfo
- Publication number
- TW201938993A TW201938993A TW108100199A TW108100199A TW201938993A TW 201938993 A TW201938993 A TW 201938993A TW 108100199 A TW108100199 A TW 108100199A TW 108100199 A TW108100199 A TW 108100199A TW 201938993 A TW201938993 A TW 201938993A
- Authority
- TW
- Taiwan
- Prior art keywords
- user
- page
- path
- current user
- historical
- Prior art date
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/44—Arrangements for executing specific programs
- G06F9/451—Execution arrangements for user interfaces
- G06F9/453—Help systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F8/00—Arrangements for software engineering
- G06F8/30—Creation or generation of source code
- G06F8/31—Programming languages or programming paradigms
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/44—Arrangements for executing specific programs
- G06F9/445—Program loading or initiating
- G06F9/44505—Configuring for program initiating, e.g. using registry, configuration files
- G06F9/4451—User profiles; Roaming
Landscapes
- Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computing Systems (AREA)
- Human Computer Interaction (AREA)
- User Interface Of Digital Computer (AREA)
Abstract
Description
本說明書一個或多個實施例是相關於電腦之技術領域,特別是相關於操作路徑導航的方法和裝置。One or more embodiments of the present specification relate to the technical field of computers, and in particular, to methods and devices for operating path navigation.
隨著網際網路技術的發展,人們常常採用各種用戶端軟體,例如手機端APP來進行各種事務處理,包括購物、支付、社交、理財等。這就使得用戶端軟體與使用者的互動設計非常重要。複雜的用戶端軟體功能繁多,如果互動設計不夠合理,使用者不能透過簡單快捷的方式抵達和執行目標操作,將會降低使用者的使用體驗,也使得產品的服務能力下降。特別是,對於一些時效性要求更高的操作,例如,一些財富管理軟體產品中基金、股票交易類的操作,如果不能快速執行目標操作甚至可能造成使用者的財產損失。並且,另一態樣,不同使用者具有不同的操作習慣和訴求。這就為用戶端軟體的使用者互動設計提出了更高的要求。
因此,希望能有改進的方案,透過有效的操作路徑導航,幫助使用者快速、高效地執行目標操作。With the development of Internet technology, people often use a variety of client software, such as mobile phone APPs, for various transactions, including shopping, payment, social networking, and financial management. This makes the interaction between the client software and the user very important. Complex client-side software has many functions. If the interaction design is not reasonable enough, users cannot reach and perform target operations in a simple and fast way, which will reduce the user experience and reduce the service capabilities of the product. In particular, for some operations that require more timeliness, such as funds and stock trading operations in some wealth management software products, if the target operation cannot be performed quickly, it may even cause the user's property loss. And, in another aspect, different users have different operating habits and demands. This puts forward higher requirements for user interaction design of client software.
Therefore, it is hoped that there can be improved solutions to help users perform target operations quickly and efficiently through effective operation path navigation.
本說明書一個或多個實施例描述了一種方法和裝置,透過操作路徑導航,幫助使用者快速執行目標操作。
根據第一態樣,提供了一種使用者操作路徑導航的方法,包括:獲取當前使用者的操作序列和使用者特徵資料;採用路徑導航模型,基於所述操作序列和使用者特徵資料,預測所述當前使用者的目標操作頁面;向所述當前使用者推薦所述目標操作頁面;根據當前使用者是否接受所推薦的目標操作頁面,確定對所述路徑導航模型的修改。
在一個實施例中,上述路徑導航模型被構建用於,基於多個使用者的歷史操作資料集和使用者特徵資料集,建立多個使用者群與操作模式的對應關係。
根據一種可能的設計,透過以下方式預測當前使用者的目標操作頁面:根據所述使用者特徵資料,確定所述當前使用者對應的特定使用者群;根據路徑導航模型建立的多個使用者群與操作模式的對應關係,確定特定使用者群對應的特定操作模式;根據所述特定操作模式,確定與所述當前使用者的操作序列匹配的特定操作路徑;根據所述特定操作路徑,確定所述當前使用者的目標操作頁面。
根據一種可能的設計,透過以下方式之一提供所述目標操作頁面:
在原有介面上提供浮層圖標,所述浮層圖標指向所述目標操作頁面;
在介面中新增圖標,該新增的圖標指向所述目標操作頁面;
將所述目標操作頁面作為下一操作頁面提供給所述當前使用者。
根據一種實施方式,確定對路徑導航模型的修改可以包括:根據當前使用者是否接受所推薦的目標操作頁面,確定所述路徑導航模型的推薦準確度;在所述推薦準確度低於預設閾值的情況下,對所述路徑導航模型進行修改。
進一步地,在一個實施例中,對路徑導航模型進行修改包括:記錄所述當前使用者的後續操作;根據所述操作序列和所述後續操作,形成目標操作路徑;根據所述目標操作路徑,更新所述路徑導航模型所建立的多個使用者群與操作模式的對應關係。
根據第二態樣,提供一種用於使用者操作路徑導航的裝置,包括:獲取單元,配置為獲取當前使用者的操作序列和使用者特徵資料;預測單元,配置為採用路徑導航模型,基於所述操作序列和使用者特徵資料,預測所述當前使用者的目標操作頁面;推薦單元,配置為向所述當前使用者推薦所述目標操作頁面;以及確定單元,配置為根據當前使用者是否接受所推薦的目標操作頁面,確定對所述路徑導航模型的修改。
根據第三態樣,提供了一種電腦可讀儲存媒體,其上儲存有電腦程式,當所述電腦程式在電腦中執行時,令電腦執行第一態樣的方法。
根據第四態樣,提供了一種計算設備,包括記憶體和處理器,其特徵在於,所述記憶體中儲存有可執行碼,所述處理器執行所述可執行碼時,實現第一態樣的方法。
透過本說明書實施例提供的方法及裝置,採用路徑導航模型向使用者推薦目標操作頁面,同時根據使用者回饋不斷訓練和改進該模型,使得模型的推薦準確度不斷提高,從而更大程度地幫助使用者快速達到目標頁面,簡化使用者操作,提升使用者體驗。One or more embodiments of the present specification describe a method and a device that help a user quickly perform a target operation by navigating through an operation path.
According to a first aspect, a user-operated path navigation method is provided, which includes: obtaining a current user's operation sequence and user characteristic data; and adopting a path navigation model based on the operation sequence and user characteristic data to predict all The target operation page of the current user is described; the target operation page is recommended to the current user; and the modification of the route navigation model is determined according to whether the current user accepts the recommended target operation page.
In one embodiment, the above-mentioned path navigation model is configured to establish a correspondence relationship between a plurality of user groups and an operation mode based on a plurality of user historical operation data sets and user characteristic data sets.
According to a possible design, the target user's target operation page is predicted in the following ways: determining a specific user group corresponding to the current user according to the user characteristic data; and multiple user groups established according to a path navigation model Correspondence with the operation mode to determine a specific operation mode corresponding to a specific user group; according to the specific operation mode, determine a specific operation path that matches the operation sequence of the current user; and determine the specific operation path according to the specific operation path Describe the target operation page of the current user.
According to one possible design, the target operation page is provided in one of the following ways:
Providing a floating layer icon on the original interface, the floating layer icon pointing to the target operation page;
Adding an icon to the interface, the added icon pointing to the target operation page;
Providing the target operation page as the next operation page to the current user.
According to an embodiment, determining the modification to the route navigation model may include: determining a recommendation accuracy of the route navigation model according to whether the current user accepts the recommended target operation page; and when the recommendation accuracy is lower than a preset threshold In the case, the route navigation model is modified.
Further, in one embodiment, modifying the route navigation model includes: recording subsequent operations of the current user; forming a target operation path according to the operation sequence and the subsequent operations; and according to the target operation path, Update the correspondence between multiple user groups and operation modes established by the route navigation model.
According to a second aspect, a device for user-guided path navigation is provided, including: an acquisition unit configured to acquire a current user's operation sequence and user characteristic data; a prediction unit configured to adopt a path navigation model based on The operation sequence and user characteristic data to predict a target operation page of the current user; a recommendation unit configured to recommend the target operation page to the current user; and a determination unit configured to be based on whether the current user accepts The recommended target operation page determines the modification of the route navigation model.
According to a third aspect, a computer-readable storage medium is provided on which a computer program is stored, and when the computer program is executed in the computer, the method for causing the computer to execute the first aspect.
According to a fourth aspect, a computing device is provided, including a memory and a processor, wherein the memory stores executable code, and the processor implements the first state when the processor executes the executable code. Kind of method.
Through the methods and devices provided by the embodiments of the present specification, a path navigation model is used to recommend a target operation page to the user, and the model is continuously trained and improved according to user feedback, so that the model's recommendation accuracy is continuously improved, thereby helping to a greater extent Users quickly reach the target page, simplifying user operations and improving user experience.
下面結合附圖,對本說明書提供的方案進行描述。
圖1為本說明書揭露的一個實施例的實施場景示意圖。如圖所示,各個使用者透過使用用戶端軟體,例如支付寶App,螞蟻財富App等,可以形成使用者歷史操作資訊。計算平台,例如上述用戶端軟體的伺服器,可以基於這些使用者歷史操作資訊,以及使用者的特徵資訊,構建一個路徑導航模型。該路徑導航模型透過分析上述歷史操作資訊和使用者特徵資訊,得出各類使用者的使用習慣。接著,可以在使用該路徑導航模型進行使用者操作路徑導航的同時,對該模型進行測試和訓練。當獲取到使用者A進入該用戶端軟體進行操作的操作序列時,將捕獲到的操作序列以及該使用者A的使用者特徵資訊輸入上述路徑導航模型。路徑導航模型基於之前對各類使用者使用習慣的分析,預測該使用者A接下來的目標操作頁面。於是,計算平台將該目標操作頁面推薦給使用者。接著,根據使用者A是否選擇了所推薦的目標操作,來判斷目標操作的預測準確度,以該預測準確度作為回饋,修改路徑導航模型。透過這樣的過程,一邊使用該模型向使用者進行目標操作的推薦,一邊根據使用者的選擇作為回饋結果來修改模型,從而不斷提高模型推薦的準確度,更好地為使用者推薦目標操作,幫助使用者高效達到目標操作。下面描述以上過程的具體實施步驟。
根據本說明描述的一種實施方式,在第一階段,針對使用者歷史操作資料集和使用者特徵資訊資料集建立路徑導航模型。在建模階段,要獲取至少兩態樣的資訊:使用者歷史操作資訊資料集,和使用者特徵資訊資料集。針對這兩態樣的資訊建立路徑導航模型,該模型用於對上述使用者歷史操作資訊和使用者特徵資訊進行分析和學習,以此得出各類使用者的操作習慣。
使用者歷史操作資訊包括使用者歷史操作軌跡,或稱為歷史操作路徑。為了獲取這樣的資訊,一般地,在使用者於終端介面上進行操作的時候,捕獲使用者的操作行為和相關頁面。使用者的操作行為可以包括使用者對頁面進行的各種操作,例如點擊、滑動、拖曳、字元鍵入、放大或縮小等等。上述頁面可以是網頁頁面,Web應用頁面,App頁面或者其他可與使用者進行互動的頁面。可以採用在終端應用埋點的方式來捕獲使用者的操作。例如,預先對頁面中的各個可互動模組,例如按鈕、滾動條、banner (橫幅廣告)等等,進行編號,以便於記錄使用者操作。
圖2示出在一個例子中使用者的操作路徑的示意圖。在圖2中,以支付寶為例,示出使用者對頁面的操作。如圖2所示,使用者從支付寶主頁面A1開始,點擊“螞蟻財富”(執行操作P1),進入頁面A2。在該A2頁面,點擊“理財”(執行操作P2),進入理財頁面A3。在A3頁面中進一步顯示了各種理財項目。假定使用者在A3頁面點擊“基金”(執行操作P3),從而進入頁面A4,其中示出各種基金選擇。在該頁面A4中,使用者又點擊“基金排行”(執行操作P4),於是進入頁面A5,在該頁面A5,透過點擊“估值排行”(執行操作P5),示出頁面A6。使用者瀏覽頁面A6的內容之後,關閉了用戶端。如此,A1.P1-A2.P2-A3.P3-A4.P4-A5.P5-A6示出一條使用者的操作路徑OP1。
在一個實施例中,在捕獲使用者操作的同時,還記錄相關的時間資訊。如此,使用者歷史操作資訊還包括操作時刻、停留時間等時間參數資訊。例如,對於圖3所示的操作路徑OP1:A1.P1-A2.P2-A3.P3-A4.P4-A5.P5-A6,還可以記錄使用者執行各個操作P1-P2-P3-P4-P5的時刻,例如T1-T2-T3-T4-T5,以及使用者在各個頁面的停留時間,例如ST1-ST2-ST3-ST4-ST5-ST6。還可以基於操作時刻和/或停留時間,確定操作的時間間隔等資訊。
在一個實施例中,以上的使用者歷史操作資訊可以記錄為序列、向量、矩陣等多種形式。例如,在一個例子中,將一條操作路徑記錄為一個向量,每個向量元素對應一個使用者操作所針對的可互動模組的編號。在一個例子中,將操作路徑,以及該操作路徑對應的時間資訊,整理為相同長度的序列,並將這些序列整理為矩陣形式。可以理解,還可以採用其他格式或形式記錄使用者歷史操作資訊。
除了使用者歷史操作資訊,另一態樣,還獲取使用者的特徵資訊。在一個實施例中,使用者特徵資訊包括,使用者的基本屬性資訊,例如註冊資訊中所包含的屬性資訊,比如年齡、性別、地域等。在一個實施例中,使用者特徵資訊還包括,與用戶端軟體功能相關的使用者的其他屬性資訊。例如,對於金融類的用戶端軟體,例如支付寶,螞蟻財富,還可以獲取使用者的餘額、基金總淨值、交易頻率等資訊作為其特徵資訊。
在一個實施例中,使用者特徵資訊包括使用者肖像資訊。使用者肖像資訊是根據使用者的基本屬性資料而抽象出的一個標籤化的使用者模型。已經存在一些方法,基於使用者的基本屬性資料(例如註冊資訊,消費記錄等),為使用者賦予一些標籤,這些標籤的集合構成使用者肖像資訊。例如,在一個具體例子中,使用者肖像資訊包括:“註冊新人”,“理財小白”,“儲蓄能手”,“中等抗風險等級”,“全職媽媽”等標籤資訊,其中“註冊新人”的標籤根據使用者註冊資訊中的註冊時間而確定,關於理財、儲蓄、抗風險能力的標籤根據使用者的例如餘額、理財收益、儲蓄等級、消費記錄等資訊綜合確定。
在一個實施例中,獲取使用者的基本屬性資料,然後採用上述方法,基於這些基本的屬性資料,確定出使用者肖像資訊。在另一實施例中,用戶端軟體出於其他功能目的,已經為使用者繪製了使用者肖像。在這樣的情況下,直接獲取這樣的使用者肖像資訊作為使用者特徵資訊。
可以理解,為了分析各類使用者群體的操作習慣,需要獲取大量使用者各自的歷史操作資訊和使用者特徵資訊。針對這兩態樣的資訊建立路徑導航模型,該模型用於對上述使用者歷史操作資訊和使用者特徵資訊進行分析和學習,以此得出各類使用者的操作習慣。
根據一個實施例,在上述路徑導航模型中採用聚類演算法,從而對大量使用者資訊進行分類分析,並提升模型的快速回歸。例如,可以對使用者特徵資訊進行聚類,從而將大量使用者按照一定粒度劃分為多個使用者群,以便為各個使用者群構建其操作行為模型。也可以對使用者歷史操作資訊進行聚類分析,從而將複雜的使用者操作行為處理為一些操作行為模式。
在聚類過程中,可以採取例如K-means的直接劃分聚類方法,基於層次的聚類演算法,例如BIRCH演算法、CURE演算法等,基於密度的聚類演算法,例如DBSCAN演算法、OPTICS演算法等等,以及其他可能的聚類演算法。
根據一個實施例,上述路徑導航模型可以採用神經網路的演算法和結構,例如卷積神經網路CNN,循環神經網路RNN等。路徑導航模型可以採用上述神經網路的演算法和結構,對輸入資料,即大量使用者的使用者歷史操作資訊和使用者特徵資訊,進行深度學習。
透過採用各種演算法,路徑導航模型基於大量使用者歷史操作資訊和使用者特徵資訊,分析和學習各個使用者群的操作使用習慣,即建立多個使用者群與操作模式的對應關係。這個分析和學習的過程具體可以包括,統計各種歷史操作路徑的頻率,操作時間,根據歷史操作路徑和操作時間參數形成多種操作模式,統計各個使用者群與操作模式的對應關係,從而為各個使用者群構建操作行為模型。更具體而言,上述使用者群可以是根據各種粒度而設定的群體,例如,在一個例子中,使用者群的劃分可以與使用者肖像中包含的人群標籤相對應一致;在另一例子中,使用者群的劃分可以對應於使用者肖像中各種人群標籤的組合,比如使用者群1對應於同時具有“理財小白”,“儲蓄能手”等標籤的使用者,使用者群2對應於同時具有“理財高手”,“股市入門”等標籤的使用者。在又一例子中,使用者群的劃分可以基於獨立的劃分規則。使用者群的劃分粒度可以根據業務需求而設置,在一個例子中,一個使用者群甚至可以只包含一個使用者。
另一態樣,操作模式可以包括頻率超過一定閾值的歷史操作路徑。進一步地,操作模式還可以包括,歷史操作路徑所對應的時間相關參數,例如操作時刻,停留時間等。
在一個具體例子中,使用者user1每天會頻繁地(例如超過一閾值,10次)執行圖2所示的操作路徑OP1:A1.P1-A2.P2-A3.P3-A4.P4-A5.P5-A6,以及另一操作路徑OP2:A1.P1-A2.P2-A3.P7-A8.P8-A9.P9-A10。這樣的資訊作為歷史操作資訊被記錄下來。路徑導航模型首先分析使用者user1的特徵資訊,將使用者user1劃分到使用者群G1。並且,路徑導航模型透過分析歷史操作資訊,特別是統計各種操作路徑的出現頻率,可以發現上述使用者user1的高頻操作路徑OP1和OP2。這樣的高頻率的操作路徑可以形成一種操作模式M1。更具體地,路徑導航模型透過分析該使用者user1的操作時間可以發現,使用者user1每天10點左右平均執行5次上述的操作路徑OP1,每天中午12點左右平均執行6次上述的操作路徑OP2,每天晚上11點左右平均執行7次上述的操作路徑OP1。基於這樣的資訊,路徑導航模型可以建立使用者user1所屬的使用者群G1與操作模式M1的對應關係,該操作模式M1中包含上述高頻操作路徑OP1和OP2。進一步地,在一個實施例中,操作模式M1還可以包括各個操作路徑的時間相關參數,例如高頻執行時段(OP1:10:00,23:00;OP2:12:00),停留時間等。
如此,路徑導航模型基於大量使用者的歷史操作資料集和使用者特徵資料集,建立多個使用者群與操作模式的對應關係。這樣構建的路徑導航模型可以用於對使用者操作進行路徑預測和導航。圖3示出根據一個實施例的使用者操作路徑導航的方法的流程圖。如圖3所示,該方法流程包括以下步驟:步驟31,獲取當前使用者的操作序列和使用者特徵資料;步驟32,採用路徑導航模型,基於所述操作序列和使用者特徵資料,預測所述當前使用者的目標操作頁面;步驟33,向所述當前使用者推薦所述目標操作頁面;步驟34,根據當前使用者是否接受所推薦的目標操作頁面,確定對所述路徑導航模型的修改。下面描述以上各個步驟的具體執行過程。
首先,在步驟31,獲取當前使用者的操作序列和使用者特徵資料。為了區別於前述建模階段的描述,下面的例子中以特定使用者,使用者A作為當前使用者進行描述。然而可以理解的是,當前使用者A可以是路徑導航模型建模階段所基於的歷史資料集中包含的歷史使用者,也可以是未包含在歷史資料集中的新的使用者。
與獲取歷史操作資訊類似的,可以透過頁面埋點等方式捕獲使用者A的操作序列SA。該操作序列SA可以僅包含一項操作,也可以包含多項連續的操作。這些操作可以是包括對頁面中可互動模組的點擊、滑動、拖曳、字元鍵入、放大或縮小等等。在一個具體例子中,在步驟31捕獲到,當前使用者A從支付寶主頁面A1開始,點擊“螞蟻財富”(執行操作P1),進入頁面A2,然後在該A2頁面,點擊“理財”(執行操作P2),進入理財頁面A3。如此形成操作序列SA:A1.P1-A2.P2-A3。在另一例子中,使用者A繼續操作,操作序列SA更新為:A1.P1-A2.P2-A3.P3-A4。
另一態樣,還獲取當前使用者A的使用者特徵資料。
在一個實施例中,獲取當前使用者A的屬性資料作為使用者特徵資料,上述屬性資料可以包括,註冊資訊中的資料,例如年齡、性別、註冊時間等,以及與用戶端主要功能相關的其他屬性資料,例如理財用戶端中的餘額、收益等資料。
在另一實施例中,將獲取的使用者A的屬性資料轉化為使用者肖像資料作為使用者特徵資料。具體地,首先獲取使用者A的各個屬性的屬性值(例如年齡、註冊時間、餘額、收益等等),根據這些屬性值確定使用者A所屬的人群標籤(例如“註冊新人”,“理財小白”,“儲蓄能手”等等),根據所述人群標籤確定使用者A的肖像資料作為所述使用者特徵資料。
在又一實施例中,用戶端軟體出於其他功能目的,已經為各個使用者繪製了使用者肖像。在這樣的情況下,直接獲取當前使用者A的肖像資料作為其使用者特徵資訊。
接著,在步驟32,採用路徑導航模型,基於捕獲的使用者A的操作序列和使用者特徵資料,預測當前使用者的目標操作頁面。如前所述,已經構建了路徑導航模型,該模型基於多個使用者的歷史操作資料集和使用者特徵資料集,建立多個使用者群與操作模式的對應關係。如此,可以採用這樣的路徑導航模型,基於所建立的對應關係,預測當前使用者A的目標操作頁面。
圖4示出根據一個實施例的預測目標操作頁面的流程圖,也就是步驟32的子步驟。如圖4所示,在步驟321,根據當前使用者的使用者特徵資料,確定該當前使用者對應的特定使用者群。
具體地,對於當前使用者A,如果路徑導航模型建模階段所基於的歷史資料集中已經包含該使用者(例如使用者A實際上等於前述的使用者user1),那麼可以直接根據建模階段的使用者群劃分,確定出使用者A所屬的使用者群,即使用者群G1。
如果歷史資料集中並不包含該使用者A,那麼可以基於使用者A的使用者特徵資料,以及路徑導航模型建模階段的使用者群劃分規則,確定使用者A所屬的使用者群。例如,建模階段的使用者群的劃分可以與使用者肖像中包含的人群標籤相對應一致,或者可以對應於使用者肖像中各種人群標籤的組合。在這樣的情況下,首先基於使用者A的使用者特徵資料獲取其使用者肖像資料,從而獲得使用者A的人群標籤,根據這些人群標籤和使用者群劃分規則,確定使用者A所屬的使用者群。在其他例子中,建模階段還可以基於其他的使用者群劃分規則。在此假定,根據使用者A的使用者特徵資料和使用者群劃分規則,確定出使用者A屬於使用者群G1。
在確定了使用者A所屬的特定使用者群的基礎上,在步驟322,根據路徑導航模型所建立的多個使用者群與操作模式的對應關係,確定該特定使用者群對應的特定操作模式。
如前所述,路徑導航模型透過分析和學習大量使用者的歷史操作資料集和使用者特徵資料集,建立了多個使用者群與操作模式的對應關係。在步驟321確定出使用者A所屬的使用者群G1的基礎上,可以透過已經建立的對應關係,確定出使用者群G1所對應的操作模式M1。
接著,在步驟323,根據上述特定操作模式,確定與當前使用者的操作序列匹配的特定操作路徑。一般地,操作模式可以包括頻率超過一定閾值的歷史操作路徑。進一步地,操作模式還可以包括,歷史操作路徑所對應的時間相關參數,例如操作時刻,停留時間等。在一個實施例中,在步驟323,獲取上述特定操作模式包含的歷史操作路徑,從中確定出與當前使用者的操作序列匹配的操作路徑作為上述特定操作路徑。上述的與當前使用者的操作序列匹配的操作路徑可以是,包含了該操作序列的操作路徑。在一個實施例中,還可以進一步考慮操作模式中包含的各個歷史操作路徑的時間相關參數。例如,在特定操作模式中有多個歷史操作路徑均包含當前操作序列的情況下,可以確定當前操作序列的當前操作時間,以及操作模式中記錄的各個歷史操作路徑的歷史執行時間,將其中歷史執行時間與當前操作時間最接近的歷史操作路徑確定為與當前操作序列匹配的特定操作路徑。
對於前述例子中的使用者群G1來說,其對應的操作模式M1至少包括歷史高頻操作路徑OP1:A1.P1-A2.P2-A3.P3-A4.P4-A5.P5-A6,以及OP2:A1.P1-A2.P2-A3.P7-A8.P8-A9.P9-A10。
在一個例子中,假定使用者A的當前操作序列SA包括,A1.P1-A2.P2-A3.P3-A4。透過將當前操作序列的元素與歷史操作路徑進行比較,可以看到,歷史操作路徑OP1包含了當前操作序列SA,因此OP1可以作為與SA匹配的特定操作路徑。
在另一例子中,假定使用者A的當前操作序列SA包括A1.P1-A2.P2-A3。可以看到,操作模式M1中包括的歷史操作路徑OP1和OP2均包含了當前操作序列SA,因此OP1和OP2均可以作為備選操作路徑。進一步地,可以參考操作模式M1中的時間相關參數,例如高頻執行時段。假定當前操作序列SA的操作時間為10:05,操作模式M1記錄的歷史執行時段包括:OP1:10:00和23:00,OP2:12:00。當前操作時間10:05更接近於歷史操作路徑OP1的高頻執行時間,因此,將OP1確定為與當前操作序列SA匹配的特定操作路徑。
接著,在步驟324,根據以上確定的特定操作路徑,確定當前使用者的目標操作頁面。具體地,將特定操作路徑中的結束頁面,作為當前使用者的目標操作頁面。在前述例子中,假定確定了OP1為特定操作路徑,那麼可以將OP1的結束頁面A6(圖2中對應於“估值排行”的頁面),作為使用者A的目標操作頁面。
如此,透過圖4所示的方式,在步驟32,採用路徑導航模型,預測當前使用者A的目標操作頁面。回到圖3,接下來,在步驟33,向當前使用者推薦所確定的目標操作頁面。
在一個實施例中,將所確定的目標操作頁面作為下一操作頁面提供給當前使用者。例如,假定已經確定特定操作路徑為OP1,OP1的結束頁面A6為使用者A的目標操作頁面,而使用者A的當前操作頁面為A3。如果使用者A在頁面A3上進行的操作,例如在頁面A3點擊“基金”(執行操作P3),繼續符合特定操作路徑OP1,那麼跳過中間頁面A4和A5,直接為使用者A顯示目標操作頁面A6。
在一個實施例中,為了避免干擾使用者的操作,在用戶端介面中新增圖標,該新增的圖標指向目標操作頁面。圖5示出根據一個實施例的新增圖標的示意圖。如圖5所示,在“支付寶”用戶端最下面一排的操作圖標中新增一圖標“快捷”,該圖標指向目標操作頁面。假定頁面A6被確定為使用者A的目標操作頁面,那麼在圖5的情況下,使用者A在當前操作序列過程中,透過點擊該圖標“快捷”,可以直接跳轉到頁面A6,省卻中間的頁面,例如A4和A5。
在另一實施例中,為了進一步減少對用戶端介面的修改,在原有介面上提供浮層圖標,所述浮層圖標指向目標操作頁面。圖6示出根據一個實施例的浮層圖標的示意圖。如圖6所示,在“支付寶”原有介面的基礎上,增加了浮層圖標“懂你所想”,該圖標指向目標操作頁面。在一個例子中,浮層圖標可以設計為半透明狀,以儘量少地影響原有介面的顯示。在前述例子中,假定頁面A6被確定為使用者A的目標操作頁面,那麼在圖7的情況下,使用者A在當前操作序列的任何頁面下,都可以透過點擊該浮層圖標“懂你所想”,直接跳轉到頁面A6,省卻中間頁面。
可以理解,還可以透過其他方式,例如推送方式,向使用者推薦所確定的目標操作頁面。
透過向使用者推薦目標操作頁面的方式,可以幫助使用者快速、高效地執行目標操作,到達目標頁面,提升使用者體驗。
同時,還可以根據使用者對推薦的目標頁面的接受情況作為回饋,進一步訓練、修改路徑導航模型,使其更符合使用者的使用習慣。也就是,在步驟34,根據當前使用者是否接受所推薦的目標操作頁面,確定對路徑導航模型的修改。
在一個實施例中,首先確定當前使用者是否接受了所推薦的目標操作頁面。在前述例子中,透過各種方式向使用者A推薦了目標操作頁面A6。如果使用者A在進入推薦的目標操作頁面A6之後,在很短的預定時間(例如1s)之內執行了退出操作,例如透過點擊其他圖標跳出該頁面,或操作返回鍵退出該頁面,或者關閉整個用戶端軟體,那麼可以認為使用者A並沒有選擇或者接受該推薦的目標操作頁面。如果使用者A在進入推薦的目標操作頁面之後,停留合理的時間,或者執行了瀏覽、進一步互動等操作,那麼可以認為使用者A選擇或者接受了該推薦的目標操作頁面。
根據當前使用者是否接受所推薦的目標操作頁面,可以確定路徑導航模型的推薦準確度。推薦準確度可以基於多個當前使用者對推薦的目標操作頁面的接受回饋而統計得到。例如,在向100個使用者推薦目標操作頁面之後,假定其中40個使用者接受了推薦,那麼可以認為推薦準確度為40%。針對每個當前使用者,可以根據其對推薦的接受情況,對推薦準確度進行累加,從而動態更新推薦準確度。例如假定當前使用者為使用者A,並且之前已經向20個使用者進行了目標操作頁面的推薦,其中9個使用者接受了推薦,推薦準確度為9/20。如果當前的使用者A接受了推薦的目標操作頁面,可以將推薦準確度更新為10/21;如果使用者A沒有接受推薦的目標操作頁面,可以將推薦準確度更新為9/21。
在所述推薦準確度低於預設閾值的情況下,可以對路徑導航模型進行修改。更具體地,可以記錄當前使用者的後續操作,即從推薦的目標操作頁面退出之後的操作。可以理解這樣的後續操作是反映該當前使用者實際目標的操作。根據該當前使用者之前的操作序列和上述後續操作,形成目標操作路徑,然後根據該目標操作路徑,更新路徑導航模型所建立的多個使用者群與操作模式的對應關係。更具體地,在一個實施例中,將上述目標操作路徑添加到歷史操作資料集中,形成更新的歷史操作資料集,並使得路徑導航模型針對更新的歷史操作資料集重新進行分析學習,從而重新建立使用者群與操作模式的對應關係。在另一實施例中,根據上述目標操作路徑,更新當前使用者(例如使用者A)所屬的使用者群(例如使用者群G1)的操作模式(例如操作模式M1)。具體地,將上述目標操作路徑作為新增的歷史操作路徑,針對當前使用者所屬的使用者群更新對歷史操作路徑的統計,從而更新該使用者群對應的操作模式的內容。
可以理解,還可以採用模型訓練過程中的一些常規演算法,例如梯度傳遞等,修改模型的其他參數,使得模型更好地回歸,並根據使用者回饋正向提升推薦準確度。
如此,透過路徑導航模型實現閉環的正向回饋。在向使用者推薦目標操作頁面的同時,不斷訓練和改進該模型,使得模型的推薦準確度不斷提高,從而更大程度地幫助使用者快速達到目標頁面,簡化使用者操作,提升使用者體驗。
根據另一態樣的實施例,還提供一種用於使用者操作路徑導航的裝置。圖7示出根據一個實施例的操作路徑導航裝置的示意性方塊圖。如圖7所示,該導航裝置700包括:獲取單元71,配置為獲取當前使用者的操作序列和使用者特徵資料;預測單元72,配置為採用路徑導航模型,基於所述操作序列和使用者特徵資料,預測所述當前使用者的目標操作頁面;推薦單元73,配置為向所述當前使用者推薦所述目標操作頁面;以及確定單元74,配置為根據當前使用者是否接受所推薦的目標操作頁面,確定對所述路徑導航模型的修改。
在一個實施例中,上述獲取單元71配置為,獲取所述當前使用者的至少一個屬性的屬性值,根據所述至少一個屬性的屬性值確定所述當前使用者所屬的至少一個人群標籤,根據所述至少一個人群標籤確定所述當前使用者的肖像資料作為所述使用者特徵資料。
根據一個實施例,上述路徑導航模型被構建用於,基於多個使用者的歷史操作資料集和使用者特徵資料集,建立多個使用者群與操作模式的對應關係。
在一個實施例中,上述預測單元72包括(未示出):第一確定模組,配置為根據所述使用者特徵資料,確定所述當前使用者對應的特定使用者群;第二確定模組,配置為根據所述多個使用者群與操作模式的對應關係,確定所述特定使用者群對應的特定操作模式;第三確定模組,配置為根據所述特定操作模式,確定與所述當前使用者的操作序列匹配的特定操作路徑;以及第四確定模組,配置為根據所述特定操作路徑,確定所述當前使用者的目標操作頁面。
進一步地,在一個實施例中,上述第三確定模組配置為:從所述特定操作模式中,確定包含所述操作序列的多個歷史操作路徑;確定所述操作序列的當前操作時間,以及所述多個歷史操作路徑各自的歷史執行時間;將歷史執行時間與所述當前操作時間最接近的歷史操作路徑確定為所述特定操作路徑。
根據一種實施方式,上述推薦單元73配置為,透過以下方式之一提供所述目標操作頁面:在原有介面上提供浮層圖標,所述浮層圖標指向所述目標操作頁面;在介面中新增圖標,該新增的圖標指向所述目標操作頁面;將所述目標操作頁面作為下一操作頁面提供給所述當前使用者。
在一個實施例中,上述確定單元74包括(未示出):準確度確定模組,配置為根據當前使用者是否接受所推薦的目標操作頁面,確定所述路徑導航模型的推薦準確度;修改模組,配置為在所述推薦準確度低於預設閾值的情況下,對所述路徑導航模型進行修改。
進一步地,在一個實施例中,所述修改模組配置為:記錄所述當前使用者的後續操作;根據所述操作序列和所述後續操作,形成目標操作路徑;根據所述目標操作路徑,更新所述路徑導航模型所建立的多個使用者群與操作模式的對應關係。更進一步地,在一個實施例中,上述修改模組還配置為:將所述目標操作路徑作為新增的歷史操作路徑,針對所述當前使用者所屬的特定使用者群更新歷史操作路徑的統計;根據更新的歷史操作路徑統計,更新所述特定使用者群對應的操作模式。
透過以上的裝置,採用路徑導航模型向使用者推薦目標操作頁面,同時根據使用者回饋不斷訓練和改進該模型,使得模型的推薦準確度不斷提高,從而更大程度地幫助使用者快速達到目標頁面,簡化使用者操作,提升使用者體驗。
根據另一態樣的實施例,還提供一種電腦可讀儲存媒體,其上儲存有電腦程式,當所述電腦程式在電腦中執行時,令電腦執行結合圖3所描述的方法。
根據再一態樣的實施例,還提供一種計算設備,包括記憶體和處理器,所述記憶體中儲存有可執行碼,所述處理器執行所述可執行碼時,實現結合圖3所述的方法。
本領域技術人員應該可以意識到,在上述一個或多個示例中,本發明所描述的功能可以用硬體、軟體、韌體或它們的任意組合來實現。當使用軟體實現時,可以將這些功能儲存在電腦可讀媒體中或者作為電腦可讀媒體上的一個或多個指令或碼進行傳輸。
以上所述的具體實施方式,對本發明的目的、技術方案和有益效果進行了進一步詳細說明,所應理解的是,以上所述僅為本發明的具體實施方式而已,並不用於限定本發明的保護範圍,凡在本發明的技術方案的基礎之上,所做的任何修改、等同替換、改進等,均應包括在本發明的保護範圍之內。The solutions provided in this specification are described below with reference to the drawings.
FIG. 1 is a schematic diagram of an implementation scenario of an embodiment disclosed in this specification. As shown in the figure, each user can form user historical operation information by using client software, such as Alipay App, Ant Wealth App, etc. A computing platform, such as the server of the above-mentioned client software, can build a path navigation model based on these user historical operation information and user characteristic information. The route navigation model obtains the usage habits of various users by analyzing the historical operation information and user characteristic information. Then, while using the route navigation model for user-guided route navigation, the model can be tested and trained. When the operation sequence in which the user A enters the client software for operation is obtained, the captured operation sequence and the user characteristic information of the user A are input into the path navigation model. The route navigation model predicts the next target operation page of the user A based on the previous analysis of various user usage habits. Then, the computing platform recommends the target operation page to the user. Then, according to whether the user A has selected the recommended target operation, the prediction accuracy of the target operation is judged, and the prediction accuracy is used as a feedback to modify the route navigation model. Through such a process, while using the model to recommend the target operation to the user, the model is modified according to the user's selection as a feedback result, so as to continuously improve the accuracy of the model recommendation and better recommend the target operation for the user. Help users reach their target operations efficiently. The specific implementation steps of the above process are described below.
According to an embodiment described in this description, in a first stage, a route navigation model is established for a user historical operation data set and a user characteristic information data set. In the modeling phase, at least two kinds of information should be obtained: the user historical operation information data set and the user characteristic information data set. A path navigation model is established for the two kinds of information, and the model is used to analyze and learn the historical user operation information and user characteristic information to obtain the operation habits of various users.
The user history operation information includes the user history operation track, or the history operation path. In order to obtain such information, generally, when a user performs an operation on a terminal interface, the user's operation behavior and related pages are captured. The user's operation behavior may include various operations performed by the user on the page, such as clicking, sliding, dragging, typing characters, zooming in or out, and so on. The above pages may be web pages, web application pages, App pages, or other pages that can interact with users. The user's operation can be captured by burying points in the terminal application. For example, each interactive module in the page, such as a button, a scroll bar, a banner (banner advertisement), and the like is numbered in advance to facilitate recording of user operations.
FIG. 2 is a schematic diagram showing a user's operation path in one example. In FIG. 2, Alipay is used as an example to show the user's operation on the page. As shown in FIG. 2, the user starts from the main Alipay page A1 and clicks “Ant Wealth” (perform operation P1) to enter the page A2. On this A2 page, click "Finance" (perform operation P2) to enter the wealth management page A3. Various financial management items are further displayed on the A3 page. Assume that the user clicks "funds" on the A3 page (performs operation P3), thereby entering a page A4, which shows various fund choices. In this page A4, the user clicks on "Fund Ranking" (perform operation P4), and then enters page A5. On this page A5, by clicking "valuation ranking" (perform operation P5), page A6 is displayed. After the user browses the content of page A6, the client is closed. In this way, A1.P1-A2.P2-A3.P3-A4.P4-A5.P5-A6 shows a user's operation path OP1.
In one embodiment, while capturing user operations, related time information is also recorded. In this way, the user's historical operation information also includes time parameter information such as operation time and dwell time. For example, for the operation path OP1 shown in FIG. 3: A1.P1-A2.P2-A3.P3-A4.P4-A5.P5-A6, it is also possible to record that the user performs each operation P1-P2-P3-P4- The time of P5, such as T1-T2-T3-T4-T5, and the user's time on each page, such as ST1-ST2-ST3-ST4-ST5-ST6. Information such as the time interval of the operation can also be determined based on the operation time and / or dwell time.
In one embodiment, the above historical operation information of the user may be recorded in various forms such as a sequence, a vector, and a matrix. For example, in one example, an operation path is recorded as a vector, and each vector element corresponds to the number of an interactive module targeted by a user operation. In one example, the operation path and the time information corresponding to the operation path are arranged into sequences of the same length, and these sequences are arranged into a matrix form. Understandably, the user's historical operation information can also be recorded in other formats or forms.
In addition to the user's historical operation information, in another aspect, the user's characteristic information is also obtained. In one embodiment, the user characteristic information includes basic attribute information of the user, such as attribute information included in the registration information, such as age, gender, region, and the like. In one embodiment, the user characteristic information further includes other attribute information of the user related to the client software function. For example, for financial client software, such as Alipay and Ant Wealth, information such as the user's balance, total fund value, and transaction frequency can also be obtained as its characteristic information.
In one embodiment, the user characteristic information includes user portrait information. User portrait information is a labeled user model that is abstracted based on the user's basic attribute data. There are already some methods based on the user's basic attribute data (such as registration information, consumption records, etc.) to give users some tags, and the collection of these tags constitutes user portrait information. For example, in a specific example, the user portrait information includes: "registered newcomer", "financial banker", "saving expert", "medium anti-risk level", "full-time mom" and other label information, including "registered newcomer" The tags of are determined according to the registration time in the user registration information, and the tags about wealth management, savings, and anti-risk ability are comprehensively determined based on the user's information such as balance, wealth management income, savings level, consumption record and other information.
In one embodiment, the basic attribute data of the user is obtained, and then the above method is used to determine the user portrait information based on the basic attribute data. In another embodiment, the client software has drawn a user portrait for the user for other functional purposes. In such a case, such user portrait information is directly acquired as the user characteristic information.
It can be understood that in order to analyze the operating habits of various user groups, it is necessary to obtain a large amount of historical operation information and user characteristic information of the users. A path navigation model is established for the two kinds of information, and the model is used to analyze and learn the historical user operation information and user characteristic information to obtain the operation habits of various users.
According to one embodiment, a clustering algorithm is used in the above-mentioned path navigation model, so as to classify and analyze a large amount of user information and improve the rapid regression of the model. For example, user characteristic information may be clustered, so that a large number of users are divided into multiple user groups according to a certain granularity, so as to construct an operation behavior model for each user group. It is also possible to perform cluster analysis on the user's historical operation information, thereby processing complex user operation behaviors into some operation behavior modes.
In the clustering process, direct clustering methods such as K-means, hierarchical clustering algorithms such as BIRCH algorithm, CURE algorithm, etc., and density-based clustering algorithms such as DBSCAN algorithm, OPTICS algorithms, etc., and other possible clustering algorithms.
According to an embodiment, the above path navigation model may adopt a neural network algorithm and structure, such as a convolutional neural network CNN, a recurrent neural network RNN, and the like. The path navigation model can use the above-mentioned neural network algorithm and structure to perform deep learning on input data, that is, user historical operation information and user characteristic information of a large number of users.
Through the use of various algorithms, the path navigation model is based on a large amount of user historical operation information and user characteristic information, and analyzes and learns the operating habits of each user group, that is, establishes the correspondence between multiple user groups and operation modes. This analysis and learning process may specifically include counting the frequency and operating time of various historical operating paths, forming a variety of operating modes based on historical operating paths and operating time parameters, and counting the correspondence between user groups and operating modes, and thus for each use Group of people to build operational behavior models. More specifically, the user group may be a group set according to various granularities. For example, in one example, the division of the user group may correspond to the crowd label included in the user ’s portrait; in another example, The division of the user group may correspond to the combination of various crowd tags in the user portrait. For example, the user group 1 corresponds to users who have tags such as "financial banker" and "saving expert", and the user group 2 corresponds to At the same time users with "financial master", "stock market entry" tags. In yet another example, the division of user groups may be based on independent division rules. The granularity of user groups can be set according to business requirements. In one example, a user group can even contain only one user.
In another aspect, the operation mode may include a historical operation path whose frequency exceeds a certain threshold. Further, the operation mode may further include time-related parameters corresponding to the historical operation path, such as operation time, dwell time, and the like.
In a specific example, the user user1 frequently (for example, exceeds a threshold, 10 times) executes the operation path OP1 shown in FIG. 2: A1.P1-A2.P2-A3.P3-A4.P4-A5. P5-A6, and another operating path OP2: A1.P1-A2.P2-A3.P7-A8.P8-A9.P9-A10. Such information is recorded as historical operation information. The route navigation model first analyzes the characteristic information of the user user1, and divides the user user1 into the user group G1. In addition, the path navigation model can find the high-frequency operation paths OP1 and OP2 of the user user1 by analyzing historical operation information, especially by counting the frequency of occurrence of various operation paths. Such a high-frequency operation path can form an operation mode M1. More specifically, by analyzing the operation time of the user user1, the path navigation model can be found that the user user1 performs the above-mentioned operation path OP1 five times at about 10 o'clock every day, and the above-mentioned operation path OP2 is performed 6 times at about noon on average every day The above-mentioned operation path OP1 is performed on average 7 times a day around 11 pm. Based on such information, the route navigation model can establish the correspondence between the user group G1 to which the user user1 belongs and the operation mode M1, which includes the above-mentioned high-frequency operation paths OP1 and OP2. Further, in one embodiment, the operation mode M1 may further include time-related parameters of each operation path, such as a high-frequency execution period (OP1: 10:00, 23:00; OP2: 12:00), a dwell time, and the like.
In this way, the route navigation model establishes a correspondence relationship between multiple user groups and operation modes based on a large number of historical operation data sets and user characteristic data sets of users. The path navigation model constructed in this way can be used for path prediction and navigation for user operations. FIG. 3 shows a flowchart of a method for user-guided route navigation according to an embodiment. As shown in FIG. 3, the method flow includes the following steps: step 31, obtaining the operation sequence and user characteristic data of the current user; step 32, using a path navigation model, predicting the location based on the operation sequence and user characteristic data The target operation page of the current user is described; step 33, the target operation page is recommended to the current user; step 34, the modification of the path navigation model is determined according to whether the current user accepts the recommended target operation page. . The specific execution process of the above steps is described below.
First, in step 31, an operation sequence and user characteristic data of a current user are acquired. In order to distinguish from the description of the foregoing modeling phase, in the following example, a specific user and user A are used as the current user. However, it can be understood that the current user A may be a historical user included in the historical data set based on the path navigation model modeling stage, or a new user not included in the historical data set.
Similar to the acquisition of historical operation information, the operation sequence SA of the user A can be captured by way of embedding a page. The operation sequence SA may include only one operation or a plurality of consecutive operations. These operations can include clicking, swiping, dragging, typing characters, zooming in or out, etc. on interactive modules in the page. In a specific example, it is captured in step 31 that the current user A starts from Alipay main page A1, clicks "Ant Wealth" (perform operation P1), enters page A2, and then on the A2 page, clicks "finance" Operation P2), enter the wealth management page A3. The operation sequence SA is thus formed: A1.P1-A2.P2-A3. In another example, the user A continues the operation, and the operation sequence SA is updated as: A1.P1-A2.P2-A3.P3-A4.
In another aspect, the user characteristic data of the current user A is also obtained.
In one embodiment, the attribute data of the current user A is obtained as the user characteristic data. The attribute data may include data in the registration information, such as age, gender, registration time, etc., and other information related to the main functions of the user terminal. Attribute data, such as the balance and income in the wealth management client.
In another embodiment, the obtained attribute data of the user A is converted into user portrait data as user characteristic data. Specifically, first obtain the attribute values (such as age, registration time, balance, income, etc.) of user A's attributes, and determine the population tags (such as "registered newcomer", "financial management" White "," saving expert ", etc.), according to the crowd tag, the portrait data of the user A is determined as the user characteristic data.
In yet another embodiment, the client software has drawn user portraits for various users for other functional purposes. In such a case, the portrait data of the current user A is directly acquired as its user characteristic information.
Next, in step 32, a path navigation model is used to predict the target user's target operation page based on the captured user A's operation sequence and user characteristic data. As mentioned above, a route navigation model has been constructed, which is based on the historical operation data set and user characteristic data set of multiple users, and establishes the correspondence between multiple user groups and operation modes. In this way, such a path navigation model can be used to predict the target operation page of the current user A based on the established correspondence relationship.
FIG. 4 shows a flowchart of a prediction target operation page, that is, a sub-step of step 32 according to an embodiment. As shown in FIG. 4, in step 321, a specific user group corresponding to the current user is determined according to the user characteristic data of the current user.
Specifically, for the current user A, if the historical data set based on the path navigation model modeling stage already includes the user (for example, user A is actually equal to the aforementioned user user1), then the user can directly The user group is divided to determine the user group to which the user A belongs, that is, the user group G1.
If the historical data set does not include the user A, the user group to which the user A belongs may be determined based on the user characteristic data of the user A and the user group division rules in the modeling stage of the route navigation model. For example, the division of the user group in the modeling phase may correspond to the crowd tags included in the user portrait, or may correspond to a combination of various crowd tags in the user portrait. In such a case, first obtain user portrait data of user A based on user characteristic data of user A, thereby obtaining a crowd label of user A, and determine the usage to which user A belongs according to these crowd labels and user group division rules. Person group. In other examples, the modeling phase may also be based on other user group division rules. It is assumed here that it is determined that the user A belongs to the user group G1 according to the user characteristic data of the user A and the user group division rule.
After the specific user group to which the user A belongs is determined, in step 322, a specific operation mode corresponding to the specific user group is determined according to the correspondence relationship between the multiple user groups and the operation modes established by the route navigation model. .
As mentioned above, the route navigation model establishes the correspondence between multiple user groups and operation modes by analyzing and learning historical user data sets and user characteristic data sets of a large number of users. After determining the user group G1 to which the user A belongs in step 321, the operation mode M1 corresponding to the user group G1 can be determined through the established correspondence relationship.
Next, in step 323, a specific operation path matching the operation sequence of the current user is determined according to the specific operation mode. Generally, the operation mode may include a historical operation path whose frequency exceeds a certain threshold. Further, the operation mode may further include time-related parameters corresponding to the historical operation path, such as operation time, dwell time, and the like. In one embodiment, in step 323, the historical operation path included in the specific operation mode is obtained, and an operation path matching the operation sequence of the current user is determined as the specific operation path. The operation path matching the operation sequence of the current user may be an operation path including the operation sequence. In one embodiment, time-related parameters of each historical operation path included in the operation mode may be further considered. For example, in a case where multiple historical operation paths in a specific operation mode all include the current operation sequence, the current operation time of the current operation sequence and the historical execution time of each historical operation path recorded in the operation mode may be determined, and the historical The historical operation path whose execution time is closest to the current operation time is determined as the specific operation path that matches the current operation sequence.
For the user group G1 in the foregoing example, the corresponding operation mode M1 includes at least the historical high-frequency operation path OP1: A1.P1-A2.P2-A3.P3-A4.P4-A5.P5-A6, and OP2: A1.P1-A2.P2-A3.P7-A8.P8-A9.P9-A10.
In one example, it is assumed that the current operation sequence SA of the user A includes, A1.P1-A2.P2-A3.P3-A4. By comparing the elements of the current operation sequence with the historical operation path, it can be seen that the historical operation path OP1 contains the current operation sequence SA, so OP1 can be used as a specific operation path that matches the SA.
In another example, it is assumed that the current operation sequence SA of the user A includes A1.P1-A2.P2-A3. It can be seen that the historical operation paths OP1 and OP2 included in the operation mode M1 both include the current operation sequence SA, so both OP1 and OP2 can be used as alternative operation paths. Further, reference may be made to time-related parameters in the operation mode M1, such as a high-frequency execution period. It is assumed that the operation time of the current operation sequence SA is 10:05, and the historical execution period recorded in the operation mode M1 includes: OP1: 10: 00 and 23:00, and OP2: 12: 00. The current operation time of 10:05 is closer to the high-frequency execution time of the historical operation path OP1. Therefore, OP1 is determined as a specific operation path that matches the current operation sequence SA.
Next, in step 324, a target operation page of the current user is determined according to the specific operation path determined above. Specifically, the end page in the specific operation path is used as the target operation page of the current user. In the foregoing example, assuming that OP1 is determined to be a specific operation path, the end page A6 of OP1 (the page corresponding to the “evaluation ranking” in FIG. 2) can be used as the target operation page of user A.
In this way, in the manner shown in FIG. 4, in step 32, a route navigation model is used to predict the target operation page of the current user A. Returning to FIG. 3, next, at step 33, the determined target operation page is recommended to the current user.
In one embodiment, the determined target operation page is provided to the current user as the next operation page. For example, it is assumed that the specific operation path has been determined to be OP1, the end page A6 of OP1 is the target operation page of user A, and the current operation page of user A is A3. If the operation performed by user A on page A3, for example, clicking "fund" (perform operation P3) on page A3, continues to conform to the specific operation path OP1, then skip the intermediate pages A4 and A5 and directly display the target operation for user A Page A6.
In one embodiment, in order to avoid disturbing the operation of the user, an icon is added to the user interface, and the newly added icon points to the target operation page. FIG. 5 shows a schematic diagram of adding an icon according to an embodiment. As shown in FIG. 5, an icon “Quick” is added to the operation icon in the bottom row of the “Alipay” client terminal, and the icon points to the target operation page. Assume that the page A6 is determined as the target operation page of the user A. In the case of FIG. 5, during the current operation sequence, the user A can directly jump to the page A6 by clicking the icon “Quick”, eliminating the intermediate Pages, such as A4 and A5.
In another embodiment, in order to further reduce the modification of the user interface, a floating layer icon is provided on the original interface, and the floating layer icon points to the target operation page. FIG. 6 shows a schematic diagram of a floating layer icon according to one embodiment. As shown in FIG. 6, on the basis of the original interface of "Alipay", a floating layer icon "Understand what you want" is added, and the icon points to the target operation page. In one example, the floating layer icon can be designed to be translucent to minimize the display of the original interface. In the foregoing example, assuming page A6 is determined as the target operation page of user A, in the case of FIG. 7, user A can click on the floating layer icon "understand you" under any page of the current operation sequence What you want "jumps directly to page A6, eliminating the intermediate page.
It can be understood that the determined target operation page can also be recommended to the user through other methods, such as a push method.
By recommending the target operation page to the user, the user can help the user perform the target operation quickly and efficiently, reach the target page, and improve the user experience.
At the same time, according to the user's acceptance of the recommended target page as feedback, the route navigation model can be further trained and modified to make it more consistent with the user's usage habits. That is, in step 34, the modification of the route navigation model is determined according to whether the current user accepts the recommended target operation page.
In one embodiment, it is first determined whether the current user has accepted the recommended target operation page. In the foregoing example, the target operation page A6 is recommended to the user A in various ways. If user A performs the exit operation within a short predetermined time (for example, 1s) after entering the recommended target operation page A6, for example, by clicking other icons to jump out of the page, or operating the return key to exit the page, or closing The entire client software, then it can be considered that User A has not selected or accepted the recommended target operation page. If user A stays for a reasonable time after performing the recommended target operation page, or performs operations such as browsing and further interaction, then user A can be considered to have selected or accepted the recommended target operation page.
According to whether the current user accepts the recommended target operation page, the recommendation accuracy of the route navigation model can be determined. The recommendation accuracy can be statistically calculated based on the feedback received by multiple current users on the recommended target operation page. For example, after recommending the target operation page to 100 users, assuming that 40 users have accepted the recommendation, the recommendation accuracy can be considered to be 40%. For each current user, the accuracy of the recommendation can be accumulated according to their acceptance of the recommendation, thereby dynamically updating the recommendation accuracy. For example, it is assumed that the current user is user A and has previously recommended the target operation page to 20 users, of which 9 users have accepted the recommendation, and the recommendation accuracy is 9/20. If the current user A accepts the recommended target operation page, the recommendation accuracy can be updated to 10/21; if the user A does not accept the recommended target operation page, the recommendation accuracy can be updated to 9/21.
In a case where the recommendation accuracy is lower than a preset threshold, the route navigation model may be modified. More specifically, a subsequent operation of the current user, that is, an operation after exiting from the recommended target operation page may be recorded. It can be understood that such subsequent operations are operations that reflect the actual goals of the current user. According to the previous operation sequence of the current user and the above-mentioned subsequent operations, a target operation path is formed, and then according to the target operation path, the correspondence relationship between multiple user groups and operation modes established by the path navigation model is updated. More specifically, in one embodiment, the target operation path is added to the historical operation data set to form an updated historical operation data set, and the path navigation model is re-analyzed and learned for the updated historical operation data set, thereby re-establishing Correspondence between user groups and operation modes. In another embodiment, according to the target operation path, the operation mode (for example, operation mode M1) of the user group (for example, user group G1) to which the current user (for example, user A) belongs is updated. Specifically, the target operation path is used as a newly added historical operation path, and statistics on the historical operation path are updated for a user group to which the current user belongs, thereby updating the content of the operation mode corresponding to the user group.
It can be understood that some conventional algorithms in the model training process, such as gradient transfer, can also be used to modify other parameters of the model to make the model return better and improve the recommendation accuracy based on user feedback.
In this way, closed-loop forward feedback is achieved through the path navigation model. While recommending the target operation page to the user, the model is continuously trained and improved, so that the recommendation accuracy of the model is continuously improved, thereby helping the user to reach the target page quickly, simplifying the user operation, and improving the user experience.
According to another aspect of the embodiment, a device for user-guided route navigation is also provided. FIG. 7 shows a schematic block diagram of an operation path navigation device according to an embodiment. As shown in FIG. 7, the navigation device 700 includes: an obtaining unit 71 configured to obtain an operation sequence and user characteristic data of a current user; and a prediction unit 72 configured to adopt a path navigation model based on the operation sequence and the user Characteristic data to predict the target operation page of the current user; a recommendation unit 73 configured to recommend the target operation page to the current user; and a determination unit 74 configured to determine whether the current user accepts the recommended target Operate the page to determine the modification to the route navigation model.
In one embodiment, the obtaining unit 71 is configured to obtain an attribute value of at least one attribute of the current user, determine at least one crowd tag to which the current user belongs according to the attribute value of the at least one attribute, and The at least one crowd tag determines the portrait data of the current user as the user characteristic data.
According to an embodiment, the above-mentioned path navigation model is constructed to establish a correspondence relationship between a plurality of user groups and an operation mode based on a plurality of users' historical operation data sets and user characteristic data sets.
In one embodiment, the prediction unit 72 includes (not shown): a first determination module configured to determine a specific user group corresponding to the current user according to the user characteristic data; a second determination module A group configured to determine a specific operation mode corresponding to the specific user group according to the correspondence between the plurality of user groups and the operation mode; a third determination module configured to determine the corresponding operation mode according to the specific operation mode The specific operation path matching the operation sequence of the current user is described; and a fourth determination module is configured to determine the target operation page of the current user according to the specific operation path.
Further, in one embodiment, the third determining module is configured to determine a plurality of historical operation paths including the operation sequence from the specific operation mode, determine a current operation time of the operation sequence, and Historical execution time of each of the plurality of historical operation paths; determining a historical operation path whose historical execution time is closest to the current operation time as the specific operation path.
According to one embodiment, the recommendation unit 73 is configured to provide the target operation page through one of the following methods: providing a floating layer icon on the original interface, the floating layer icon pointing to the target operation page; adding a new interface to the interface Icon, the newly added icon points to the target operation page; the target operation page is provided as the next operation page to the current user.
In one embodiment, the above-mentioned determining unit 74 includes (not shown): an accuracy determining module configured to determine the recommended accuracy of the route navigation model according to whether the current user accepts the recommended target operation page; modify A module configured to modify the route navigation model when the recommended accuracy is lower than a preset threshold.
Further, in one embodiment, the modification module is configured to: record subsequent operations of the current user; form a target operation path according to the operation sequence and the subsequent operations; and according to the target operation path, Update the correspondence between multiple user groups and operation modes established by the route navigation model. Furthermore, in an embodiment, the modification module is further configured to: use the target operation path as a newly added historical operation path, and update statistics of historical operation paths for a specific user group to which the current user belongs. ; Update the operation mode corresponding to the specific user group according to the updated historical operation path statistics.
Through the above devices, the path navigation model is used to recommend the target operation page to the user, and the model is continuously trained and improved according to user feedback, so that the model's recommendation accuracy is continuously improved, thereby helping users to quickly reach the target page. , Simplifying user operations and improving user experience.
According to another aspect of the embodiment, a computer-readable storage medium is further provided, on which a computer program is stored, and when the computer program is executed in the computer, the computer is caused to execute the method described in conjunction with FIG. 3.
According to yet another aspect of the embodiment, a computing device is further provided, which includes a memory and a processor. The memory stores executable code. When the processor executes the executable code, the combination is implemented in combination with FIG. 3. The method described.
Those skilled in the art should appreciate that, in one or more of the above examples, the functions described in the present invention may be implemented by hardware, software, firmware, or any combination thereof. When implemented in software, these functions may be stored in a computer-readable medium or transmitted as one or more instructions or codes on a computer-readable medium.
The specific embodiments described above further describe the objectives, technical solutions, and beneficial effects of the present invention in detail. It should be understood that the above are only specific embodiments of the present invention and are not intended to limit the present invention. The scope of protection, any modification, equivalent replacement, or improvement made on the basis of the technical solution of the present invention shall be included in the scope of protection of the present invention.
A1‧‧‧頁面A1‧‧‧Page
A2‧‧‧頁面 A2‧‧‧ page
A3‧‧‧頁面 A3‧‧‧ page
A4‧‧‧頁面 A4‧‧‧ pages
A5‧‧‧頁面 A5‧‧‧ page
A6‧‧‧頁面 A6‧‧‧ pages
P1‧‧‧操作 P1‧‧‧operation
P2‧‧‧操作 P2‧‧‧operation
P3‧‧‧操作 P3‧‧‧ Operation
P4‧‧‧操作 P4‧‧‧ Operation
P5‧‧‧操作 P5‧‧‧operation
31‧‧‧步驟 31‧‧‧step
32‧‧‧步驟 32‧‧‧ steps
33‧‧‧步驟 33‧‧‧step
34‧‧‧步驟 34‧‧‧step
321‧‧‧步驟 321‧‧‧step
322‧‧‧步驟 322‧‧‧step
323‧‧‧步驟 323‧‧‧step
324‧‧‧步驟 324‧‧‧step
700‧‧‧導航裝置 700‧‧‧Navigation device
71‧‧‧獲取單元 71‧‧‧Get Unit
72‧‧‧預測單元 72‧‧‧ Forecast Unit
73‧‧‧推薦單元 73‧‧‧Recommended Unit
74‧‧‧確定單元 74‧‧‧ confirm unit
為了更清楚地說明本發明實施例的技術方案,下面將對實施例描述中所需要使用的附圖作簡單地介紹,顯而易見地,下面描述中的附圖僅僅是本發明的一些實施例,對於本領域普通技術人員來講,在不付出創造性勞動的前提下,還可以根據這些附圖獲得其它的附圖。In order to explain the technical solution of the embodiment of the present invention more clearly, the drawings used in the description of the embodiments are briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. Those of ordinary skill in the art can obtain other drawings according to the drawings without paying creative labor.
圖1為本說明書揭露的一個實施例的實施場景示意圖; FIG. 1 is a schematic diagram of an implementation scenario of an embodiment disclosed in this specification; FIG.
圖2示出在一個例子中使用者的操作路徑的示意圖; FIG. 2 shows a schematic diagram of an operation path of a user in an example; FIG.
圖3示出根據一個實施例的使用者操作路徑導航的方法的流程圖; 3 shows a flowchart of a method for user-guided path navigation according to an embodiment;
圖4示出根據一個實施例的預測目標操作頁面的流程圖; 4 shows a flowchart of a prediction target operation page according to an embodiment;
圖5示出根據一個實施例的新增圖標的示意圖; FIG. 5 illustrates a schematic diagram of adding an icon according to an embodiment; FIG.
圖6示出根據一個實施例的浮層圖標的示意圖;以及 Figure 6 shows a schematic diagram of a floating layer icon according to one embodiment; and
圖7示出根據一個實施例的操作路徑導航裝置的示意性方塊圖。 FIG. 7 shows a schematic block diagram of an operation path navigation device according to an embodiment.
Claims (20)
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810220308.5A CN108647064A (en) | 2018-03-16 | 2018-03-16 | The method and device of courses of action navigation |
??201810220308.5 | 2018-03-16 |
Publications (1)
Publication Number | Publication Date |
---|---|
TW201938993A true TW201938993A (en) | 2019-10-01 |
Family
ID=63744217
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
TW108100199A TW201938993A (en) | 2018-03-16 | 2019-01-03 | Operation path navigation |
Country Status (3)
Country | Link |
---|---|
CN (1) | CN108647064A (en) |
TW (1) | TW201938993A (en) |
WO (1) | WO2019174379A1 (en) |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108647064A (en) * | 2018-03-16 | 2018-10-12 | 阿里巴巴集团控股有限公司 | The method and device of courses of action navigation |
CN110032597B (en) * | 2018-11-30 | 2023-04-11 | 创新先进技术有限公司 | Visual processing method and device for operation behaviors of application program |
CN109492831B (en) * | 2018-12-20 | 2023-12-29 | 平安科技(深圳)有限公司 | Marketing path recommendation method and device, medium and electronic equipment |
CN110750718A (en) * | 2019-10-10 | 2020-02-04 | 深圳Tcl新技术有限公司 | User guiding method, device, equipment and storage medium based on recommendation system |
CN111767465A (en) * | 2020-07-09 | 2020-10-13 | 中国工商银行股份有限公司 | Information recommendation method and device, computing equipment and medium |
Family Cites Families (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070121653A1 (en) * | 2005-11-04 | 2007-05-31 | Reckamp Steven R | Protocol independent application layer for an automation network |
CN106294614B (en) * | 2016-08-01 | 2019-09-20 | 华为技术有限公司 | Method and apparatus for accessing business |
CN107728874A (en) * | 2017-09-06 | 2018-02-23 | 阿里巴巴集团控股有限公司 | The method, apparatus and equipment of user prompt operation are provided |
CN108647064A (en) * | 2018-03-16 | 2018-10-12 | 阿里巴巴集团控股有限公司 | The method and device of courses of action navigation |
-
2018
- 2018-03-16 CN CN201810220308.5A patent/CN108647064A/en active Pending
-
2019
- 2019-01-03 TW TW108100199A patent/TW201938993A/en unknown
- 2019-01-07 WO PCT/CN2019/070617 patent/WO2019174379A1/en active Application Filing
Also Published As
Publication number | Publication date |
---|---|
CN108647064A (en) | 2018-10-12 |
WO2019174379A1 (en) | 2019-09-19 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11044949B2 (en) | Systems and methods for dynamic delivery of web content | |
KR102408476B1 (en) | Method for predicing purchase probability based on behavior sequence of user and apparatus therefor | |
WO2019174379A1 (en) | Operation path navigation | |
US20180374108A1 (en) | Method and Apparatus for Building a User Profile, for Personalization Using Interaction Data, and for Generating, Identifying, and Capturing User Data Across Interactions Using Unique User Identification | |
Zhang et al. | Price does matter! modeling price and interest preferences in session-based recommendation | |
CN110309119B (en) | System, method and apparatus for implementing data upload, processing and predictive query API disclosure | |
KR20200123015A (en) | Information recommendation method, apparatus, device and medium | |
Miao et al. | Context‐based dynamic pricing with online clustering | |
Lytvyn et al. | Design of a recommendation system based on Collaborative Filtering and machine learning considering personal needs of the user | |
US20170091847A1 (en) | Automated feature identification based on review mapping | |
Singh et al. | How to maximize clicks for display advertisement in digital marketing? A reinforcement learning approach | |
US20190228105A1 (en) | Dynamic website content optimization | |
Lytvyn et al. | Design of the architecture of an intelligent system for distributing commercial content in the internet space based on SEO-technologies, neural networks, and Machine Learning | |
AU2015317621A1 (en) | Method and apparatus for predicting customer intentions | |
US10678831B2 (en) | Page journey determination from fingerprint information in web event journals | |
CN109324862B (en) | APP display method and device | |
Kumar et al. | A computational framework for ranking prediction of cloud services under fuzzy environment | |
Ghazal | Fischer linear discrimination and quadratic discrimination analysis–based data mining technique for internet of things framework for healthcare | |
US11676183B1 (en) | Translator-based scoring and benchmarking for user experience testing and design optimizations | |
US12039424B2 (en) | Systems and methods for managing, distributing and deploying a recursive decisioning system based on continuously updating machine learning models | |
US20170140428A1 (en) | Method and apparatus for dynamically selecting content for online visitors | |
Ballandies et al. | Mobile link prediction: Automated creation and crowdsourced validation of knowledge graphs | |
US11615077B1 (en) | Editable rules-based model system | |
US11080555B2 (en) | Crowd sourced trends and recommendations | |
US20230186376A1 (en) | Systems and methods for user interface orchestration and presentation |