"time series prediction"

Vikas · August 2010

Hi everyone

I am new user of RapidMiner and this is my first post, I have 11 months electric feeder load time series data so I want to forecast one day ahead feeder load with the help of this data.so can anyone guide me how can I do this with the help of RapidMiner ?

Data Format:-
date hour1 hour2 hour3 hour4 hour5 hour6 ......... hour24
10/01/2010 .2934 .1983 .1328 .2032 .1002 .1834 ......... .2903
10/02/2010 .2367 . 1298 .1289 .1901 .1192 .1920 ........ .1902
................. ................................................................................
280 days 24 hour

Thanks

Vikas Gupta

wessel · August 2010

First convert your data to:

day_time, load
1, .2934
2, .1983
....
n .1902

Then use the windowing operator with the appropriate embedding dimension.
Then use k-nn or linear regression as a learner.

If you upload like 50 rows of data I'll make you an example process.

IngoRM · August 2010

Dear Vikas,

please post your questions only once in the most appropriate board and not in every board here. Thanks.

Cheers,
Ingo

wessel · August 2010

Actually I have no idea how to convert this data to 1 column, using rapidminer, so I'm gonna use VIM or python.

Like:
loaddata #<-- this is a comment
0.5144 #<-- first data point
0.5144
0.5144
0.6001
0.6001
0.6859
0.6859
0.7716
0.7716
1.286
1.286
1.286
1.2003
1.2003
1.2003
1.286
1.286
1.5432
1.8004
1.6289
1.5432
1.3717
1.1145
0.8573
0.9431 #<-- day 2
...
1.286 #<-- last data point, etc

edit:
okay here is the data I will be using:

Vikas · August 2010

Thanks for help me Wessel

Please help me about windowing operator(horizon,window size) to forecast the feeder load one day ahead.

Thanks
Vikas

wessel · August 2010

Your problem does not seem to be really interesting.
So your better of with classical statistics. No need for windowing, embedding, and machine learning here.

http://devio.us/~wessel/load/load.jpeg
http://devio.us/~wessel/load/load2.jpeg

Vikas · August 2010

Actually I am trying to build a model where I can predict the load in advance(1 or 2 day ahead) with the help of previous load data which can improve load shedding management of electric feeder.

wessel · August 2010

Meh, if you insist creating a model using heavy number crunching machine learning....

here is the process:

<?xml version="1.0" encoding="UTF-8" standalone="no"?>
<process version="5.0">
  <context>
    <input/>
    <output/>
    <macros/>
  </context>
  <operator activated="true" class="process" compatibility="5.0.9" expanded="true" name="Process">
    <parameter key="logfile" value="/home/wessel/loaddata.aml"/>
    <process expanded="true" height="507" width="705">
      <operator activated="true" class="read_csv" compatibility="5.0.9" expanded="true" height="60" name="Read CSV" width="90" x="45" y="30">
        <parameter key="file_name" value="/home/wessel/Desktop/loaddata.csv"/>
        <parameter key="column_separators" value="   "/>
        <parameter key="date_format" value="MM/dd/yyyy"/>
        <list key="data_set_meta_data_information"/>
      </operator>
      <operator activated="false" class="series:moving_average" compatibility="5.0.2" expanded="true" height="76" name="MA_3_cen" width="90" x="180" y="30">
        <parameter key="attribute_name" value="load"/>
        <parameter key="window_width" value="3"/>
        <parameter key="result_position" value="center"/>
        <parameter key="keep_original_attribute" value="false"/>
      </operator>
      <operator activated="false" class="rename" compatibility="5.0.9" expanded="true" height="76" name="Rename" width="90" x="315" y="30">
        <parameter key="old_name" value="moving_average(load)"/>
        <parameter key="new_name" value="ma3_load"/>
      </operator>
      <operator activated="true" class="filter_examples" compatibility="5.0.9" expanded="true" height="76" name="Filter Examples" width="90" x="450" y="30">
        <parameter key="condition_class" value="no_missing_attributes"/>
      </operator>
      <operator activated="true" class="series:windowing" compatibility="5.0.2" expanded="true" height="76" name="Windowing" width="90" x="585" y="30">
        <parameter key="horizon" value="24"/>
        <parameter key="window_size" value="24"/>
        <parameter key="create_label" value="true"/>
        <parameter key="label_attribute" value="load"/>
      </operator>
      <operator activated="false" class="principal_component_analysis" compatibility="5.0.9" expanded="true" height="94" name="PCA" width="90" x="45" y="120">
        <parameter key="dimensionality_reduction" value="fixed number"/>
        <parameter key="number_of_components" value="4"/>
      </operator>
      <operator activated="true" class="optimize_selection" compatibility="5.0.9" expanded="true" height="94" name="Optimize Selection" width="90" x="180" y="120">
        <parameter key="generations_without_improval" value="2"/>
        <parameter key="keep_best" value="2"/>
        <process expanded="true" height="507" width="784">
          <operator activated="true" class="series:sliding_window_validation" compatibility="5.0.2" expanded="true" height="112" name="Validation" width="90" x="186" y="163">
            <parameter key="training_window_width" value="24"/>
            <parameter key="test_window_width" value="24"/>
            <parameter key="horizon" value="24"/>
            <parameter key="cumulative_training" value="true"/>
            <process expanded="true" height="507" width="165">
              <operator activated="true" class="linear_regression" compatibility="5.0.9" expanded="true" height="94" name="Linear Regression" width="90" x="45" y="30">
                <parameter key="feature_selection" value="none"/>
              </operator>
              <connect from_port="training" to_op="Linear Regression" to_port="training set"/>
              <connect from_op="Linear Regression" from_port="model" to_port="model"/>
              <portSpacing port="source_training" spacing="0"/>
              <portSpacing port="sink_model" spacing="0"/>
              <portSpacing port="sink_through 1" spacing="0"/>
            </process>
            <process expanded="true" height="507" width="300">
              <operator activated="true" class="apply_model" compatibility="5.0.9" expanded="true" height="76" name="Apply Model" width="90" x="45" y="30">
                <list key="application_parameters"/>
              </operator>
              <operator activated="true" class="performance_regression" compatibility="5.0.9" expanded="true" height="76" name="Performance" width="90" x="45" y="120">
                <parameter key="root_mean_squared_error" value="false"/>
                <parameter key="correlation" value="true"/>
              </operator>
              <connect from_port="model" to_op="Apply Model" to_port="model"/>
              <connect from_port="test set" to_op="Apply Model" to_port="unlabelled data"/>
              <connect from_op="Apply Model" from_port="labelled data" to_op="Performance" to_port="labelled data"/>
              <connect from_op="Performance" from_port="performance" to_port="averagable 1"/>
              <portSpacing port="source_model" spacing="0"/>
              <portSpacing port="source_test set" spacing="0"/>
              <portSpacing port="source_through 1" spacing="0"/>
              <portSpacing port="sink_averagable 1" spacing="0"/>
              <portSpacing port="sink_averagable 2" spacing="0"/>
            </process>
          </operator>
          <connect from_port="example set" to_op="Validation" to_port="training"/>
          <connect from_op="Validation" from_port="averagable 1" to_port="performance"/>
          <portSpacing port="source_example set" spacing="0"/>
          <portSpacing port="source_through 1" spacing="0"/>
          <portSpacing port="sink_performance" spacing="0"/>
        </process>
      </operator>
      <operator activated="true" class="select_by_weights" compatibility="5.0.9" expanded="true" height="94" name="Select by Weights" width="90" x="315" y="120"/>
      <operator activated="true" class="series:sliding_window_validation" compatibility="5.0.2" expanded="true" height="112" name="Validation2" width="90" x="450" y="120">
        <parameter key="training_window_width" value="24"/>
        <parameter key="test_window_width" value="24"/>
        <parameter key="horizon" value="24"/>
        <parameter key="cumulative_training" value="true"/>
        <process expanded="true" height="507" width="300">
          <operator activated="true" class="linear_regression" compatibility="5.0.9" expanded="true" height="94" name="LinearR2" width="90" x="180" y="30">
            <parameter key="feature_selection" value="none"/>
          </operator>
          <connect from_port="training" to_op="LinearR2" to_port="training set"/>
          <connect from_op="LinearR2" from_port="model" to_port="model"/>
          <portSpacing port="source_training" spacing="0"/>
          <portSpacing port="sink_model" spacing="0"/>
          <portSpacing port="sink_through 1" spacing="0"/>
        </process>
        <process expanded="true" height="507" width="300">
          <operator activated="true" class="apply_model" compatibility="5.0.9" expanded="true" height="76" name="ApplyM2" width="90" x="45" y="30">
            <list key="application_parameters"/>
          </operator>
          <operator activated="true" class="write_aml" compatibility="5.0.9" expanded="true" height="60" name="Write AML" width="90" x="60" y="160">
            <parameter key="example_set_file" value="/home/wessel/loaddata.dat"/>
            <parameter key="attribute_description_file" value="/home/wessel/loaddata.aml"/>
          </operator>
          <operator activated="true" class="performance_regression" compatibility="5.0.9" expanded="true" height="76" name="Perf2" width="90" x="45" y="300">
            <parameter key="root_mean_squared_error" value="false"/>
            <parameter key="correlation" value="true"/>
          </operator>
          <connect from_port="model" to_op="ApplyM2" to_port="model"/>
          <connect from_port="test set" to_op="ApplyM2" to_port="unlabelled data"/>
          <connect from_op="ApplyM2" from_port="labelled data" to_op="Write AML" to_port="input"/>
          <connect from_op="Write AML" from_port="through" to_op="Perf2" to_port="labelled data"/>
          <connect from_op="Perf2" from_port="performance" to_port="averagable 1"/>
          <portSpacing port="source_model" spacing="0"/>
          <portSpacing port="source_test set" spacing="0"/>
          <portSpacing port="source_through 1" spacing="0"/>
          <portSpacing port="sink_averagable 1" spacing="0"/>
          <portSpacing port="sink_averagable 2" spacing="0"/>
        </process>
      </operator>
      <operator activated="true" class="read_aml" compatibility="5.0.9" expanded="true" height="60" name="Read AML" width="90" x="447" y="300">
        <parameter key="attributes" value="/home/wessel/loaddata.aml"/>
        <parameter key="column_separators" value=" "/>
      </operator>
      <connect from_op="Read CSV" from_port="output" to_op="Filter Examples" to_port="example set input"/>
      <connect from_op="Filter Examples" from_port="example set output" to_op="Windowing" to_port="example set input"/>
      <connect from_op="Windowing" from_port="example set output" to_op="Optimize Selection" to_port="example set in"/>
      <connect from_op="Optimize Selection" from_port="example set out" to_op="Select by Weights" to_port="example set input"/>
      <connect from_op="Optimize Selection" from_port="weights" to_op="Select by Weights" to_port="weights"/>
      <connect from_op="Select by Weights" from_port="example set output" to_op="Validation2" to_port="training"/>
      <connect from_op="Select by Weights" from_port="original" to_port="result 3"/>
      <connect from_op="Validation2" from_port="model" to_port="result 1"/>
      <connect from_op="Validation2" from_port="averagable 1" to_port="result 4"/>
      <connect from_op="Read AML" from_port="output" to_port="result 2"/>
      <portSpacing port="source_input 1" spacing="0"/>
      <portSpacing port="sink_result 1" spacing="144"/>
      <portSpacing port="sink_result 2" spacing="0"/>
      <portSpacing port="sink_result 3" spacing="0"/>
      <portSpacing port="sink_result 4" spacing="0"/>
      <portSpacing port="sink_result 5" spacing="0"/>
    </process>
  </operator>
</process>

wessel · August 2010

Result:
correlation: 0.795 +/- 0.136 (mikro: 0.786)

Vikas · August 2010

If we see the data there are many outlier(0 and equal load) or human intervention so for better result should I perform outlier analysis before the forecasting ?

wessel · August 2010

I don't know, it depends on your application.

A correlation of 0.8 is already really good.

Depends also on how much noise your sensor has.

wessel · August 2010

So, ehm, you got the "process" to run?

Vikas · August 2010

Dear Wessel

I got the process but please give me some help abut it's Output

1:- Prediction trend accuracy and correlation both are same thing?
2:-Can you give me some explanation about its output of process?

Thanks
Vikas

wessel · August 2010

No they are not the same, but they are related.

Look at the scatter plot of predicted load vs actual load.
When a data point is predicted correctly it lies exactly on the diagonal.
You see that all data points that are not 0 are predicted with only a small error.

The error is bigger in data points that are 0, which is expected because they are anomalous values.

I could have used "mean absolute error" instead of "correlation".
But the nice thing about "correlation" is that its invariant to the dataset.
If I would multiply all data points by a factor 100, "mean absolute error" would go up by a factor 100.
Correlation stays the same, since its normalized between -1 and 1.

Vikas · August 2010

Hi Wessel

Can you help me about this linear regression generated by process
0.299 * load-23 - 0.041 * load-19 + 0.006 * load-15 - 0.007 * load-8 - 0.014 * load-5 + 0.217 * load-1 + 0.407 * load-0 + 0.182
for forecasting of one day ahead load.

wessel · August 2010

What you want to know about this?

Vikas · August 2010

this is regression equation so how can I forecast(calculate) of load at 12,11,10,7 hour can you show me one example?

wessel · August 2010

I'm not sure I understand what you are asking.

Vikas · August 2010

Can you suggest me any other alternative for prediction of (one day load) with the help of previous model ?

wessel · August 2010

An alternative would be: Arima:
http://en.wikipedia.org/wiki/Autoregressive_integrated_moving_average

Vikas · August 2010

Can I apply ARIMA using RapidMiner operators for forecasting the load?

wessel · August 2010

No.

Vikas · August 2010

Hi Wessel

Can you give me some idea about one hour ahead load prediction using same data set ?

Thanks

wessel · August 2010

I gave you a full implementation in Rapid Miner, and a link to alternative approach,...

isn't that ideas enough?

"time series prediction"

Answers

Categories