Menu

Issue in API - Others Also Bought

Chithra
2015-08-05
2015-08-26
  • Stephan Zavrel

    Stephan Zavrel - 2015-08-05

    Hi!

    Since your integration looks fine, are you sure the rule calculation was executed?
    Usually new recommendations are calculated only once a day (default during the night at 2 am). You can also trigger rule generation manually using the puzzle piece symbol next to your tenant in the Administration section.

    If after plugin execution there are still no recommendations, it may be that the action you sent to easyrec so far did not lead to any conclusive results for the algorithm. But this is difficult to say without further knowledge of your action distribution.

    Greetings
    Stephan

     

    Last edit: Stephan Zavrel 2015-08-05
    • Chithra

      Chithra - 2015-08-06

      I have tried to access this many days. So I guess the rule calculation was executed.
      Please let me know what other information you need to debug furthur?

       
      • Stephan Zavrel

        Stephan Zavrel - 2015-08-06

        Hi!

        Can you take a look at the Plugin Logs in the Administration section. Do you see log output of the plugins there (there should be entries for ervery day? Take a look at the output generated and it can give you a clue as to why there are no results (number of actions considered, number of baskets,...). Maybe you can post the output of one log here so we can have a look at it.

        Greetings
        Stephan

         
  • Chithra

    Chithra - 2015-08-07

    Please see the log of "BOUGHT_TOGETHER"

    Configuration:

    <armConfiguration> <configurationName>Default Configuration</configurationName> <associationType>BOUGHT_TOGETHER</associationType> <tenantId>1</tenantId> <itemTypes/> <actionType>BUY</actionType> <supportPrcnt>0.0</supportPrcnt> <supportMinAbs>2</supportMinAbs> <confidencePrcnt>0.0</confidencePrcnt> <maxRulesPerItem>50</maxRulesPerItem> <excludeSingleItemBaskests>false</excludeSingleItemBaskests> <ratingNeutral>5.5</ratingNeutral> <metricType>CONFIDENCE</metricType> <maxSizeL1>5000</maxSizeL1> <doDeltaUpdate>false</doDeltaUpdate> </armConfiguration>

    Statistics:

    <armStatistics> <endDate>2015-08-07T02:00:40.744+05:30</endDate> <startDate>2015-08-07T02:00:40.727+05:30</startDate> <numberOfActionsConsidered>0</numberOfActionsConsidered> <numberOfRulesCreated>0</numberOfRulesCreated> <duration>0</duration> <lastConf>0.0</lastConf> <lastSupport>2</lastSupport> <metricType>CONFIDENCE</metricType> <nrBaskets>0</nrBaskets> <nrProducts>0</nrProducts> <sizeCountMap>0</sizeCountMap> <sizeL1>0</sizeL1> <sizeL2>0</sizeL2> <sizeRules>0</sizeRules> </armStatistics>
     
  • Florian Kleedorfer

    As it says 'numberOfActionsConsidered' is 0 in the armStatistics, are you sure that any BUY actions have actually been sent to easyrec prior to the execution of the plugin?

    One way to check is to look at the 'overview' section for the teanant in your admin interface - It looks to me like you shouldn't see any buy actions there.

     
  • Chithra

    Chithra - 2015-08-07

    Hi,
    Buy actions are there in the overview, Please see the attached screenshot

     
  • Florian Kleedorfer

    You won't get any recommendations if every buy is made by a different user, or if all buys are made by the same user (I believe). Might one of these be the case?

     
  • Chithra

    Chithra - 2015-08-10

    By different user, do u mean easyrec user or the commerce store user??

     
  • Florian Kleedorfer

    I meant different commerce store users.

     
  • Chithra

    Chithra - 2015-08-10

    Hi,

    Updating another log below:
    s
    <?xml version="1.0" encoding="UTF-8" standalone="yes"?>
    <armStatistics>
    <endDate>2015-08-10T02:00:09.805+05:30</endDate>
    <startDate>2015-08-10T02:00:09.723+05:30</startDate>
    <numberOfActionsConsidered>179</numberOfActionsConsidered>
    <numberOfRulesCreated>0</numberOfRulesCreated>
    <duration>0</duration>
    <lastConf>0.0</lastConf>
    <lastSupport>2</lastSupport>
    <metricType>CONFIDENCE</metricType>
    <nrBaskets>4</nrBaskets>
    <nrProducts>23</nrProducts>
    <sizeCountMap>0</sizeCountMap>
    <sizeL1>17</sizeL1>
    <sizeL2>0</sizeL2>
    <sizeRules>0</sizeRules>
    </armStatistics>

    In this numberOfActionsConsidered is not zero. But still we are not getting any recommendations on 'usersalsobought'. Can you please check if we are doing anything wrong?

    REST API invoked : http://localhost:8080/easyrec-web/api/1.0/otherusersalsobought?apikey=8ab9dc3ffcdac576d0f298043a60517a&tenantid=InfyGroceryStore&itemid=GFR033_3302&userid=64006&itemtype=ITEM&requesteditemtype=ITEM
    where 64006 is the user id stored in our DB

     
  • Stephan Zavrel

    Stephan Zavrel - 2015-08-10

    Hi!

    Given the output you posted it looks like the data you send into the system is still too sparse for the algorithm to find any commonalities in user behaviour. You need more data from more different users. If the actions don't show any common behaviour easyrec cannot deliver any recommendations because there are simply no predictions to be made.

    The absolute minimum action set for the ARM plugin to find recommendations looks as follows:

    For two users (A,B) and three items (1,2,3):

    A buys 1
    B buys 1
    A buys 2
    B buys 2
    A buys 3
    B buys 3

    Now you have 2 shopping baskets A(1,2,3) and B(1,2,3). It can now be concluded that if a user buys product 1, he/she is very likely to buy(like) product 2 or 3 as well.

    Try to send actions as described above to your system and start the rule calculation. After that you should get recommendations for items 1, 2 and 3 and have proof that your integration is set up correctly. It is then just the distribution of your actions that does not lead to any conclusive results yet.​

     
  • Chithra

    Chithra - 2015-08-13

    Hi Stephen,

    We tried the same approach. But still we are not able to get the recommendations.
    However the "mostvieweditems" and "mostboughtitems" api yield results.

     
  • Florian Kleedorfer

    Hi,

    Easyrec removes items from recommendations that were previously seen/bought by the user (to avoid showing the same items over and over), which might lead to your situation (no recommendations) when there are only a few actions, as is the case when there are only a few 'test users' in the system.
    To see if this is the case, please have a look at the admin tool and check if you see items that have rules of type BOUGHT_TOGETHER. If there are, they are not shown through the API because of the filter.

    In any case, please post the plugin log again.

    Best,
    Florian

     
  • Chithra

    Chithra - 2015-08-26

    BOUGHT_TOGETHER:
    <?xml version="1.0" encoding="UTF-8" standalone="yes"?>
    <armStatistics>
    <endDate>2015-08-26T02:00:35.173+05:30</endDate>
    <startDate>2015-08-26T02:00:35.150+05:30</startDate>
    <numberOfActionsConsidered>0</numberOfActionsConsidered>
    <numberOfRulesCreated>0</numberOfRulesCreated>
    <duration>0</duration>
    <lastConf>0.0</lastConf>
    <lastSupport>2</lastSupport>
    <metricType>CONFIDENCE</metricType>
    <nrBaskets>0</nrBaskets>
    <nrProducts>0</nrProducts>
    <sizeCountMap>0</sizeCountMap>
    <sizeL1>0</sizeL1>
    <sizeL2>0</sizeL2>
    <sizeRules>0</sizeRules>
    </armStatistics>

    VIEWED_TOGETHER:
    <?xml version="1.0" encoding="UTF-8" standalone="yes"?>
    <armStatistics>
    <endDate>2015-08-26T02:00:34.905+05:30</endDate>
    <startDate>2015-08-26T02:00:34.109+05:30</startDate>
    <numberOfActionsConsidered>6</numberOfActionsConsidered>
    <numberOfRulesCreated>6</numberOfRulesCreated>
    <duration>0</duration>
    <lastConf>0.0</lastConf>
    <lastSupport>2</lastSupport>
    <metricType>CONFIDENCE</metricType>
    <nrBaskets>2</nrBaskets>
    <nrProducts>3</nrProducts>
    <sizeCountMap>0</sizeCountMap>
    <sizeL1>3</sizeL1>
    <sizeL2>3</sizeL2>
    <sizeRules>6</sizeRules>
    </armStatistics>

     

Log in to post a comment.

Want the latest updates on software, tech news, and AI?
Get latest updates about software, tech news, and AI from SourceForge directly in your inbox once a month.