Create a Time Series Processing Model
1. What is Time Series Processing?
Time series prediction refers to an artificial intelligence model that identifies past patterns and predicts what values the data will have in the future by learning historical data that changes over time.
It is a model that learns how B affects A with the values you want to predict (A) and other factors (B) that affect it, and the relationship with A the next day, and predicts future A values through it.
It is a model that is mainly used to predict market prices of certain items such as energy consumption, stock forecasting, gold or oil prices.
To classify categories by combining two or more data, you can find more information in Combining Data.
Try it yourself
Learning Objective: To Identify patterns of past energy usage to predict energy usage Create an AI model
1. View Data
Data features date and power consumption data per hour are taken from the PJM website, consisting of the usage values of energy in megawatts (MW).
The result is PJME_MW, which predicts the energy value after 1 hour.
2. Create an AI Model
1) Upload data
1-1) Click DS2 DATASET for data upload.
1-2) Click Add Data to add data for artificial intelligence development.
1-3) Because downloaded data is a .csv file, click CSV and click Next.
1-4) Click the Find File button to click and upload the data you want to download.
1-5) You can check the uploaded file and select the result column in the data settings.
1-6) Data is being uploaded.
2) Develop artificial intelligence
2-1) Select your training data to create AI model at DS2 DATASET and click START AI button .
Please choose the learning type and preferred method for the project.
2-2) Learning type - Select 'Time series prediction' for the learning type.
2-3) Analysis Criteria - The preferred method is divided into two, and I'll choose the 'high accuracy' method.
2-4) Analysis Units - Set analysis units to fit the purpose of the model in months, days, hours, minutes to learn. If you want to select an analysis unit based on the data type, please select 'Automatic'.
2-5) At the bottom, you can choose whether to use learning data or not along with data summarization. (Values that you want to analyze/predict will automatically be disabled.)
2-6) Preprocess Data → You can preprocess by selecting the value you want to preprocess.
- If necessary, click the preprocess button to perform the preprocessing.
2-7) Select the desired preprocessing features and click Finish.
2-8) Completed preprocessing can be verified by marking completion.
2-9) After all the work is done, click Start in the top right to create artificial intelligence.
2-10) creating artificial intelligence.
3. View/Predict AI Details
CLICLICK AI can select the optimal AI model by generating multiple AI models and checking the model-specific details. Once all AI models are created, Error Rate can be checked and compared for each model, and detailed views, individual predictions, and batch predictions can be made for each model.
1-1) Model performance evaluation
In Detail, you can check the Error Rate and MASE of the model.
1-2) Feature Importance
Feature Importance is the importance of a variable, indicating how much influence each variable has on its predictions.
1-4) Sharing a service app
If you click the "Share service app" button to the right of "TRAINING MODEL (model number)", you can predict and analyze outside of DS2.AI.
2) Single prediction
If you select Single Prediction, you can get one prediction value as a result by entering the value of each variable. It can be used to test the accuracy of a single prediction or model. You can check the result by entering data that matches the key.
3) Collective prediction
3-1) If you select Batch Prediction, you can predict a lot of data at once. Click Download Predictive Template to download the template. If the letters in the downloaded csv file are broken, please refer to the encoding conversion.
3-2) Populate the data into the template you downloaded.
3-3) Upload the predictive template that filled the data, then click Next.
3-4) The batch forecast begins, and you can receive the batch forecast results through notification.
3-5) Once the batch prediction is complete, you can check it through the notification window.
4. Take advantage of artificial intelligence
It can be used as a form of download, deployment, and sale for each AI model generated.
- You can purchase a model license by clicking Download. When purchasing a model license, email the Deep Learning model file with code that enables Inference functionality on Jupyter.
*If you purchase Jetson Nano 2GB Developer Kit's chipset separately, you can use the artificial intelligence model through embedded products without connecting to the operating server.
- When you choose a deployment, you rent a cloud server to provide integrated MLops to distribute, operate, and manage that model.
- Select the cloud provider and region you want to see the list of available instances.
- Select the desired instance and click "CREATE CLUSTER" at the bottom to take advantage of AIOps linked to [SKYHUB AI | Deploy].
If you choose to sell, you can sell the generated model to the AI Market of DS2.AI. Please select the desired price and sales option of the model and click on the sales request. It will be uploaded to AI Market after reviewing the adequacy of AI model and data security procedures.
Models that are determined to be available for sale can be found on the AI Market's product list, and if the customer's model is purchased by other users who need it, it generates a set amount of revenue (20% commission).