Develop and Deploy Machine Learning Models With Entendu

Entendu is a web app allowing users to develop and deploy popular machine learning algorithms through a simplified user interface.

Here’s what you can do with entendu:

  • Train machine learners on your data to create custom predictive models.
  • Use those models through auto-generated forms.
  • Do bulk predictions on csv data.
  • Give model access to your clients or staff.

Our development server is now open for a small beta test. (We are a small team and we will likely close the beta to new accounts when we reach a few beta testers).

Do you want to develop a predictive algorithm with your data? We welcome you to join the open beta!

Visit entendu.io to signup.

Feel free to email me directly about the project at [email protected]

Thank you for your help!

Entendu Demonstration:

In this document:

  1. Create a free Entendu account
  2. Set up a model
  3. Train a model
  4. Using your model
  5. Assigning the model
  6. Proper formatting
    • Exporting from excel
    • Exporting from google sheets
    • Exporting from mac numbers
    • Exporting from postgresql
    • Exporting from oracle sql developer
    • Exporting from sql server
    • Exporting from mongoDb

Step 1: Create a free Entendu account

  • Navigate to https://entendu.io/register
  • Register (Make sure you have access to this email. If your password is ever lost, this email will be used for password recovery.)
  • Login with the email and password you used to create the account.

Step 2: Set up your Model

  • Navigate to the Create page using the site navigation.
  • Drop a formatted csv file into the drop area.
  • Confirm the data looks correct using the Data Preview Table.
  • Choose the label. This is the column you would like your model to predict.
  • Columns to exclude. Exclude data that is not valuable in making this prediction. Ex: customer numbers, customer names, emails
  • Training split: The machine learner will use a percentage of your data to train and a percentage of your data to test itself. A recommended split is 80/20 (.8 in this example)
  • Choose a learning algorithm. Take into consideration the label you are predicting.
    • If you are predicting a continuous number such as the price of a home, this is a regression predictor.
    • If you are predicting a categorical result such as low, medium, or high, this is a categorical predictor.
    • XGBoost or CatBoost. These are both Gradient Boosting Algorithms that handle learning in slightly different ways. One is not always better than the other. Intricacies in the data will likely cause one of them to be the better predictive algorithm. This short conversation is a good high level overview of each: convo
    • Click get parameters to return your training parameters form.

Step 3: Train your model

  • Each machine learning algorithm has its own parameters. The parameters have help dialogs to explain what the parameter tunes.
  • Choose your parameters.
    • The loss function can be set to auto or a function can be chosen. In XGBoost this is called Objective and the learning algorithm will choose this automatically.
    • Start with a lower amount of iterations/n_estimators and depth/max_depth. This will allow you to tune parameters quickly and improve the model before doing long training runs.
  • Click Build, wait for the results to return training accuracy data and training visualizations.
  • Use the data and visualizations to adjust parameters until the model is predicting at a satisfactory accuracy/loss rate
  • When satisfied, click Save.

Step 4: Using your model

  • Start on the manage page. Here you can:
    • Name your model
    • Suspend your model (keep it but make it unusable)
    • Delete your model
    • View your model training documents
    • View the generated form
  • Click User View in the top right corner of the page. Here you will see what a user who is assigned this model will see.
  • From User View, click the arrow button next to your model to bring up its interactive form.
    • Using the form will allow you to predict using your model.
    • Fill out the form. Form fields can be left empty and the model will use the mean if it is a number field or the mode if it is a categoric field. (note that 0 is not empty. Remove the number to leave it empty)
    • Click the Submit button and the prediction will be made.
  • Click the Batch Predict button in the top right.
    • This form will batch predict given a csv file with the same column names it was trained on.
    • The csv file you upload can contain extra fields that were excluded from your original model, but it must include all of the original columns except for the label (predicted column).
    • The information excluded from training, such as customer numbers, emails and names will never be sent to the server. However, it will be combined with the prediction data to help you parse through the prediction information.
    • This information can be downloaded by clicking the button Export CSV.

Step 5: Assigning the model

  • As a data engineer, you might want to create models specifically for different clients or teams in your company.
  • Start on the User page.
  • Here you can create, edit and delete users.
    • Add a user by clicking the Add User button in the top right of the screen.
    • To edit a user click the edit icon in the users row under the Actions column.
    • To delete a user click the delete icon in the users row under the Actions column.
  • After you’ve created a user, you can assign that user model access.
    • Click the edit icon in the users Action column.
    • You will see two tables. The users Model Access table, and the Assign Models table.
    • The Assign Models table contains all of your models. To assign model access to the user, click the plus icon in the actions column.
    • To revoke model access, click the delete button in the users Model Access table.
  • When models are assigned, the user has access to use the model to make predictions. If the model is set to “suspended”, it cannot be used by any users.

Step 6: Proper formatting

Entendu works with csv files. The first row must be the column names. Each row after must be the corresponding values for each column.

View a sample here

Excel:

  • Go to File > Save As. The Save As dialog box appears.
  • In the Save as type box, choose the text file format for the worksheet. CSV (Comma delimited).
  • Browse to the location where you want to save the new text file, and then click Save.

More information: Excel CSV Info

Google Sheets:

  • Open the spreadsheet in Google Sheets.
  • Click on the tab you want to save as CSV.
  • Go to File > Download > Comma Separated Values (.csv).

More information: Google Sheets CSV Info

Mac Numbers:

  • Open the spreadsheet
  • Choose File > Export To > CSV from the File menu at the top of your screen.

More information: Mac Numbers CSV Info

Postgresql:

  • Write a copy query with – delimiter ‘,’ csv.
  • Example: COPY (select id,name from tablename) TO <‘filepath/name.csv’> DELIMITER ‘,’ CSV HEADER;

More information: Postgresql CSV Info

Oracle SQL Developer:

  • Login and Connect Your Database
  • Write a Query to Retrieve the Specific Data
  • Access the Export Wizard
  • Choose the Source & the Format
  • Review & Export the Data

More information: Oracle SQL Developer CSV Info

SQL Server:

  • In SQL Server Management Studio, connect to a database you want to export a table from
  • Right-click the database and navigate to Tasks > Export Data
  • In the SQL Server Import and Export Wizard window, click Next
  • Customize the data in the Choose a Data Source window:

    • Select SQL Server Native Client 11.0 from the Data source drop-down menu.
    • By default, the Server name and Database fields already contain appropriate data.
    • Select a required mode in the Authentication block.
    • Click Next.
  • Customize the data in the Choose a Destination window:

    • Select Flat File Destination from the Destination drop-down menu.
    • Enter the file name in the File Name field.
    • To select the file destination path, click Browse, select the path, and create the .csv file.
    • Click Next.
  • Select a required option in the Specify Table Copy or Query window and click Next.

  • Select the table you want to export from the Source table or view drop-down menu and click Next.

  • In the Save and Run Package window, click Next.

  • Read the information in the Complete the Wizard window and click Finish.

More information: SQL Server CSV Info

MongoDb:

  • Write a mongoexport statement.
  • Example: mongoexport –db database_name –collection collection_name –type=csv –fields fields_names –out path_or_name_of_the_file

More information: MongoDb CSV Info


Discover more from Coursity

Subscribe to get the latest posts sent to your email.

Leave a Comment

Your email address will not be published. Required fields are marked *

Discover more from Coursity

Subscribe now to keep reading and get access to the full archive.

Continue reading

Scroll to Top