I'm working with the MyMediaLite's item recommendation tool and I'm trying to understand how can I do in order to generate a prediction file which consists only of test set's items.
Actually with the default options (--overlap-items) it generates a prediction file which contains both items that are present in the test set and someone which aren't present in it. This is really annoying simply because I need to obtain correct recommendation in order to generate the metrics for my recommender (I use an external tool in order to generate the metrics).
So I've tried the option "--in-test-items" which should use as candidate items only those present in the test set, but this is not the result that I obtain. Actually I'm working with the movielens 100k dataset which I've binarized following a specific strategy: I've assigned 1 to items which have as rating 4 or 5, otherwise 0.
I want to report here the command that I use in order to produce the recommendation in a implicit feedback situation(supposing that I'm trying to execute the test on the first split of the dataset):
item_recommendation --training-file=u1.base --test-file=u1.test --prediction-file=u1.mml_res --in-test-items --recommender=ItemKNN --predict-items-number=10
There is something incorrect in this command? Do I have to fix something?
Thank you in advance.
Alessandro Suglia
The question was answered in this e-mail thread: https://groups.google.com/forum/#!topic/mymedialite/Sm6RXehqrYE
Related
I have searched and searched and have not been able to find the answer to this. I'm no stranger to SSRS, .Net (c# and vb.net), SQL, etc...been in it for years. I currently have a multi-select report parameter that is populated by a dataset in my report. There are hundreds of entries, so I built it to be driven by a wildcard character in a preceding parameter. Everything works fine right now. My question is this: is it possible to enter a wildcard value, select one (or more) of the filtered values and then store that/those value(s) on selection so that a user can go back and enter another wildcard value and select from a newly filtered list? (Basically, remember what has been selected in the overall dataset before report execution and create some sort of comma-separated list as the final parameter value to be passed to the report) I realize this may be better served in a web app w/a reportviewer control, but I'm trying to avoid deviating from the native SSRS server if possible. Thanks in advance!
The way I might approach this (not actually done it but the theory sounds ok)
Have 2 parameters for user input, your current one and a hidden one called say #filter (visible) and #filterHistory (this is the hidden one)
Have a textbox (formatted like button) with something like "Refine" as the text. Set the action to call your report again but set the #filterHistory to be something like #filterHistory & ", " & #filter. Basically we keep appending the last user input to the history.
Then your report would filter based on both parameters. You'll have to do some parsing of the delimited parameter now to split it out into the constituent parts but you get the idea.
I've no time to build a test report but hopefully that will point you in the right direction. If it doesn't help or work then comment and I'll see if I can knock up a quick example.
my following problem is, that I have a List of Items and want to index those with elasticsearch. I have a running elasticsearch instance, and this instance has an index called "default".
So I'm running following code:
var items = GetAListOfItem();
var response = Client.IndexMany(items);
I also tried it with Client.IndexManyAsync(items). But that didn't do anything.
Only 1 Item of this List gets indexed. Nothing more. I think its the last item, which got indexed.
I thought it could be a thing with IEnumerable and multiple enumerations, but i parsed it as a List<Item>.
Another Question would be about the best practice with Elasticsearch. Is it common to use a Index per Model. So if I'm gathering data from for example Exchange and another system, I would do 2 indeces?
ExchangeIndex
OtherSystemIndex
Thank you for your help.
Update: I saw that my Client.Index does all those calls succesful, but all those objects got the same ID from NEST. Normally she had to increment by herself, isnt it?
Update 2: I fixed the Indexing Problem. I had setup an empty ID-Field.
But still have the question mit best practive about Elasticsearch.
If you are uploading all the data with the same id, it will not increment the id, that will update the record with that id and you will have only one record, so you can upload the data without an id or give wherever unique id to identified the records.
The other common problem is that your records have not the same mapping that you give for the index.
About the other question, in the indexes, you store the information that is relevant for you, even if that have content from many models, the only thing that you have to avoid is mix information, if you have an index about server logs dont mix it with user activities for example.
I would like to know if anyone knows if it is possible to run a SAP User Compare from c# using ERPConnect 4 from Theobald? If so, how?
I can open a connection to SAP and run functions - just don't know how to do User Compare.
EDIT:
It seems like we have to run the report PFCG_TIME_DEPENDENCY.
If anyone knows how to run a report with ERPConnect, or if there exists a functional module in SAP that can run a report, that will also help.
I am not exactly sure what your comparison has to include, but I assume, that you want to compare attributes of the users. If that is the case, you could download the users data from the SAP tables. Here is a starting point for what tables you probably need: http://www.tcodesearch.com/sap-tables/detail?id=USR01
USER01 is the user master record, containing all user with it's main attributes. You can find other interesting related user table through the link above.
To read a table using Erpconnect, look at this link: https://my.theobald-software.com/index.php?/Knowledgebase/Article/View/21/23/reading-sap-tables-directly
You need to create an instance of the ReadTable class. Then you add the fields you are interested in using the AddField method (e.g. MANDT and BNAME for the USR01 table). You could but don't have to enter filter criteria using the AddCriteria method. If you do add multiple creteria, be sure to add boolean operators like "and" or "or":
table.AddCriteria("LANGU = 'D'");
table.AddCriteria("AND MANDT = '007'");
Finally set the table name of the table you want to download and execute the Run-Method. After that you can loop through the results stored in <your RunTable-Instance>.Result.Rows
Sascha
I'm new to both UniData and Uniobjects so if I ask something that obvious I apologize.
I'm trying to write a tool that will let me export contacts from our ERP (Manage2000) that runs on UniData (v. 6.1) and can then import them into AD/Exchange.
The primary issue I'm having is that I don't know which fields (columns?) in the table (file?) are for what. I know that that there is a dictionary that has this information in it but I'm not sure how to get what I want out of it.
I found that there is a command LIST.METADATA in the current UniData documentation from Rocket but it seems that either the version of UniData that we are using is so old that it doesn't have this command in it or it was removed from the VOC file for some unknown reason.
Does anyone know how or have any tips to pull out the structure of a table so that I can know which fields are for what data?
Thanks in advance!
At TCL:
LIST DICT contact.master
Please note that the database file name (EX: contact.master) is case sensitive. I don't have a UniData instance at the moment to provide an example output. However, it should be similar to Universe's output:
Field......... Type & Field........ Conversion.. Column......... Output Depth &
Name.......... Field. Definition... Code........ Heading........ Format Assoc..
Number
AMOUNT.WEBB A 1 MR22 Amt WEBB 10R M
PANDAS.COST A 3 MD2Z Pandass Cost 10R M
CREDIT.EXP.DT A 6 D4/ Cred Exp Date 10R M
For the example above, you can generally tell the "data type" of the field by looking at the conversion code. "D4/" is the conversion code for a date. "MD2Z" is a numeric conversion code, which we can guess is for monetary amounts. I'm glossing over the power of conversion codes, so please make sure to reference Rocket's documentation for these codes to truly understand what these fields would output. If you don't have the documentation in front of you, you can also reference this site:
http://www.koretech.com/kr_help/KU2/30/en/KMK_Prog_Conversions.htm
If you wanted to use UniObjects and C# to retrieve the field names in a file, you could use the following code:
UniCommand fieldSelectCommand = activeSession.CreateUniCommand();
fieldSelectCommand.Command = "SELECT DICT contact.master";
fieldSelectCommand.Execute();
UniSelectList resultList = activeSession.CreateUniSelectList(0);
String[] allFieldNames = resultList.ReadListAsStringArray();
Having answered your question, I would also like to make a recommendation that you check out Rocket's U2 Toolkit for .NET if you're mostly going to be selecting data from the database instead of reading and manipulating individual records:
http://www.rocketsoftware.com/products/rocket-u2-toolkit-net
Not only does it present an ADO.NET way of accessing the database, it also has a better performance version of the UniObjects library under the U2.Data.Client.UO namespace.
The Dictionary, in my opinion, is a recommendation of how the schema should behave. However, there are cases when it's not 100% accurate. You could run "LIST CONTACT.MASTER TOXML TO MYFILE.XML" which would create an xml file what you could parse.
See https://u2devzone.rocketsoftware.com/accelerate/articles/u2-xml/u2-xml#section-0 for more information.
I am trying to figure out the best way to match items on a datagridview to items in an access database. (Think Quicken match transaction)
I import an excel sheet into a datagridview,from there it checks the access db looks for a match - if a match is found then it reports match in a column if not unmatched is reported.
i have tried to count the rows on an sql query - if = 1 then match is yes, but that for some reason will goof up sometimes.
so i am looking for the best way to do this.
Thanks - please let me know if you need any additional info.
There isn't a simple answer to this, and it depends on what your data looks like, and what you consider a "match" to be. As a very basic answer, this is one way to attack the problem. How far you take it is up to you...
Create an algorithm that takes all fields for a row and generates a "key" for it. For example if there are two fields [First], [Last] then perhaps the key would be "Bubba|Gump"
Apply that algorithm to both sets of data (the datagrid records and the access db records).
Compare the two sets of keys to determine what's identical/missing/added.
It's not foolproof but with some additional sophistication it'll take you surprisingly far.