Regression unable to use polynomial label(or any label)

green_duck

Hello all,

new here and new to RM(which will be made obvious shortly). So, i'm trying to do a simple regression analysis based on a attribute(label) as(-1,0,1). I've followed the steps provided to me, but every time I input a regression operator, I get an error saying the operator cannot handle polynomial or numerical labels. I'm stumped.

Any help would be greatly appreciated! Thanks!

lionelderkrikor

@green_duck,

In attached file, the working process.
How said previously, you have a classification problem, thus you need a classifier model (Here I used aNaive Bayesmodel).
TheLinear Modelyou used is dedicated to regression task(s) and thus raised an error in your case.

To go further and to find the best model for your use case, I advice you to use theAuto-Modeltool : Click onAuto-Model, submit your data, choosePredictand select your label attribute (in your case "sentiment") and then follow the indications.

Good luck !

Hope this helps,

Regards,

Lionel

lionelderkrikor

@green_duck,

I'm not specialist of Deep-learning but as said previously, you can start with theDeep learningmodel proposed inAuto-Modelwith the default parameters.

Don't forget to enable the optionTurn into Classification.

You will obtain a first performance. Then you can play with the structure of the neural network by adding or removing hidden layer(s), and/or playing with the number of epochs (increase this parameter), modify the activation function etc. and see if you can improve the performance of your process.


Hope this helps,

Regards,

Lionel

lionelderkrikor

Hi@green_duck,

If your attribute(label) has (-1,0,1) as values, it is a classification problem and not a regression problem.
A regression problem is characterized by a continuous attribute(label).
Can you provide your process and your data in order we can fix your error ?

Regards,

Lionel

green_duck

Hi Lionel,

Thanks for getting back to me - I had a feeling this may have been the case as I was also attempting to use cross-validation but I couldn't get the operator to work either. I've attached the data(should've done this earlier).

XML下面:

<关键参数="logverbosity" value="init"/>

<关键参数="random_seed" value="2001"/>

<关键参数="send_mail" value="never"/>

<关键参数="notification_email" value=""/>

<关键参数="process_duration_for_mail" value="30"/>

<关键参数="encoding" value="SYSTEM"/>

<关键参数="repository_entry" value="data/Tweets_sequence"/>

<关键参数="attribute_filter_type" value="all"/>

<关键参数="attribute" value="sentiment"/>

<关键参数="attributes" value="1_word|2_word|3_word|4_word|5_word|6_word|7_word|8_word|9_word|10_word|11_word|12_word|13_word|14_word|15_word|16_word|17_word|18_word|19_word|20_word|21_word|22_word|23_word|24_word|25_word|26_word|27_word|28_word|29_word|30_word"/>

<关键参数="use_except_expression" value="false"/>

<关键参数="value_type" value="attribute_value"/>

<关键参数="use_value_type_exception" value="false"/>

<关键参数="except_value_type" value="time"/>

<关键参数="block_type" value="attribute_block"/>

<关键参数="use_block_type_exception" value="false"/>

<关键参数="except_block_type" value="value_matrix_row_start"/>

<关键参数="invert_selection" value="false"/>

<关键参数="include_special_attributes" value="true"/>

<关键参数="attribute_filter_type" value="single"/>

<关键参数="attribute" value="sentiment"/>

<关键参数="attributes" value=""/>

<关键参数="use_except_expression" value="false"/>

<关键参数="value_type" value="numeric"/>

<关键参数="use_value_type_exception" value="false"/>

<关键参数="except_value_type" value="real"/>

<关键参数="block_type" value="value_series"/>

<关键参数="use_block_type_exception" value="false"/>

<关键参数="except_block_type" value="value_series_end"/>

<关键参数="invert_selection" value="false"/>

<关键参数="include_special_attributes" value="false"/>

<关键参数="attribute_name" value="sentiment"/>

<关键参数="target_role" value="label"/>

<关键参数="ratio" value="0.8"/>

<关键参数="ratio" value="0.2"/>

<关键参数="sampling_type" value="shuffled sampling"/>

<关键参数="use_local_random_seed" value="false"/>

<关键参数="local_random_seed" value="1992"/>

<连接from_op = "选择属性”from_port = "前女友ample set output" to_op="Numerical to Polynominal" to_port="example set input"/>

<关键参数="feature_selection" value="M5 prime"/>

<关键参数="alpha" value="0.05"/>

<关键参数="max_iterations" value="10"/>

<关键参数="forward_alpha" value="0.05"/>

<关键参数="backward_alpha" value="0.05"/>

<关键参数="eliminate_colinear_features" value="true"/>

<关键参数="min_tolerance" value="0.05"/>

<关键参数="use_bias" value="true"/>

<关键参数="ridge" value="1.0E-8"/>

<关键参数="create_view" value="false"/>

<关键参数="use_example_weights" value="true"/>

green_duck

@lionelderkrikor

Thank you so much! This was very helpful - Just have one last question - are there any deep learning models(NNs) that you would suggest for this same dataset?