46527

Instructions on Using the tool

(Buitiinga predction Model)

Sfep I Enter Your Dafa

(A) Enter your data in The Data works he et. starting from the celi AC 105 (B The obs ervations s hould be in rows and the yariables s hould be in columns.

(C i Above each column. choos e appropriate Type(Omrt, Output. Cont. Cat)

To drop a column from model - s et the type = Omit To treat a column as categorical Input. s et type = C iii To treat a column as continuous Input, s et type = Cont To treat a column as Output, s et vpe = Outpu t

You can have atmostlO output yt-rr-ble.:. Application will automaticaJly treat them all as contim UsuaJly one builds prediction model with 1 output only.

If you have s ay. 2 output variables Y1 and Y2. both of which depend on the s ame s et of Input v< you may be better off. building2 separate modeb - One with Y1 as Output, another one with Y2

You can have atmost5Q input yariables. out of which atmos 14Q could be categorical

Make s ure thatthe number of lnput(C at 6 C ont) columns exacUy match with the number enter<

i D Pleas e make s ure thatyour data does not h«we blank rows or blank columns.

(E) Continuous Inputs:

Any non-number in Cont column will be treated as missing vaJue.

Application will replace it by the column mean

(E) Categorical Inputs

Any blank celi or cells containing Excel error in Cat column will be treated as missing Application will reaplce it by the most frequently occuring category.

Categoiy labek are -- .n; en- r>r./e - lables good, Good. GoOd. GOOD will all be tr< Theres hould be le a; 12 obs er/atiore in each category of a Cat column.

If one of the category of a Cat column has only 1 obs eryation. you s hould do one of R em cwe that o bs e ivation OR

Ren ame the category to any other categories of that Cat column.

Sfep 2: Fili u p Model Inputs

(A) F ill up the model inputs in the Us er Input Page.

i B Make surę that your inputs axe withinthe rangę ofyalues allowed by the application.

(C ; Click the'Build Modef button to start modeling.

Step 3: Fesu/ts of Modeling

(A A NeuraJ Network model is bas ically as et of weighk between the layers of the net.

Atthe end of the run. the finał set of weights aresaved inthe Calcsheet. i B The output page of this file will show youthevalues of MSE and ARE on the training and va as the training ofthe model progresses. Two charts showingtraining and VaJidation MSE‘s hawe been aJreadyprovided inthe Outputsheet.

i'C) In Userlnputpage if you hawe as ked to s ave the model in aseparate file. then a new file will be created containing the model inputs. your data and the fitted model ( i.e. the weights) You will be able to usethis file as a caJculatorto do prediction. gk/en any new input.

Sfep 4: Study Profiles

Fitted model t> a surface in p-dimension where the number of your inputs is p.