Description
Pr.3.
You are given hourly temperature data from 5:00AM to 1:00PM for 3 days. Using this data, predict the
temperature of the 4th day (hint: you can use your code from project 2).
The data is given in hourly increments in four text files (please download “Pr3_data.zip”). The training
data for the first 3 days are named “train_data_X.txt” where “X” is the date. Use this data for training.
Once the decision unit is trained, predict the temperatures for the 4th date for hourly increments from
6:00AM to 1:00PM. Use the testing data to calculate the error of the prediction. The testing data is in the
file “test_data_4.txt”. (Note: do not use this data for training, the testing data is only for testing.)
1) What are the inputs and outputs for this problem?
2) What should be the activation function of your decision unit? Why did you choose it? (Hint: it
should output continuous real values linearly from –INF to +INF, not asymptotically ending in 0
and 1;)
3) Train the decision unit on the data from the first 3 days. Report training error. (Clearly show in a
table the desired output and the output of your decision unit and the errors)
4) Predict the temperature of the 4th date. Calculate the error of your prediction using the data for the
4th date. Report testing error. (Clearly show in a table the desired output and the output of your
decision unit and the errors)
5) Report the number of iterations, the learning rate, and any data pre-processing steps you chose
(hint: consider normalization of input data). Clearly explain why you selected these values.
6) Based on the learning trend of the decision unit, would you be able to predict if the decision unit
would improve its prediction if more iterations were allowed?
7) Could the error be further reduced using a neural network (opposed to a single decision unit)? If
so, discuss how and why these methods would reduce the error.
Compile your answers into a single PDF file and submit along with your code.
——————————————————————
Note:
1. Your software must be user friendly. The TA must be able to test it simply by executing the code.
2. Hint: you can consider normalization of input data (denormalization of output data).
3. Project deliverable should be a zip file containing:
a. Written report with answers to the questions above in word, pdf, ps, or txt format
b. The data and separation lines in format as specified by Project1_data.zip
c. The source code.
4. Submit your zip file to Instructor mmanic@vcu.edu and cc TA Darshini (Samantha) Mahendran,
mahendrand@vcu.edu. Please use the subject line [CMSC 409] Family name, Project 3.
