No-Show Prediction

No-Show Prediction


No-show is a term referring to the act of not showing up for a reservation or an appointment without properly cancelling it. No-show cases often occur across various industries like restaurants, hospitals, and beauty salons. It causes economic and employment loss for many individuals working in the industries. Previous hospital no-show data analysis and future no-show predictions could become a potential solution for the no-show problems.
 - Predict  individuals' no-show based on hospital appointment data.
 - Discover which aspect has the biggest impact on individuals' no-shows.

Create Model

1. Check Data

[new_medical.csv]
 - This dataset is composed of information such as age, gender, illnesses, SMS, and no-show results of individual patients.
 - The target variable 'No-show' has two classes: 'Yes' and 'No'.
 - PatiendID and AppointmentID columns will not be used to train the model.

2. Build Model |  Create a customer no-show prediction model using CLICK AI's structured data category classification

2.1 Model Development

[Main Page → Click on 'Create a new project.']

['Adding data' → Upload data file to Click AI.]

[Select uploaded data file from the dataset list → Click on 'Start AI.']

2.2 Model Configuration

 [new_medical.csv]
 - The target variable 'No-Show' has two values of 'Yes' and 'No'. Select 'Structured data classification' as your learning type.
 - To develop a model focused on maximizing its accuracy, select 'with higher accuracy' as your preferred way.
 - Select 'No-show_new_medical' as your value you want to predict.
 - Click on (?) for more details about each options.

2.3 Data Preprocessing

['Summary' → Decide which columns from the data set you would like to use to train the model.]
 - The columns will be automatically unchosen from being used as part of the training data.

[Data Preprocessing → Decide which columns you would like to preprocess.]
 - If you would like to preprocess data, continue by clicking on 'Preprocessing.'

[Choose preferred preprocessing methods and click 'confirm.']

['Data' → View your data file.]

2.4 Model Generation

[After all of the above steps are completed, press the 'Start' button to begin processing the AI model.]

[All of the processed models will appear in the list at the bottom of the page. You may sort the models according to RMSE, errorRate or Mase.]

3. View Details | Identify important factors for your model

 - Let's find out the factors that have the biggest impact on the predictions.

[Training Model 28 → Click on 'Details.']
 - View details of the [Training Model 28] with the highest accuracy.

[Feature Importance → See which indices are important for the prediction model. Additional information are available throughout the different tabs.]

[ eXplainable AI ]
✓ Whether patients received SMS was analyzed to have the biggest impact on patients' no-shows. Age and hospital location also displayed significant impacts on the predictions.
 ✓ You may explore through additional details and statistics on the 'Details' page.

[API Connection→ You can connect model results through API connections.]
  - API sources are available for multiple programming languages.

4. Analyze Data |  Analyze features to derive insights from data

[Training Model 28 → Click on 'Analyze']
 - View data analysis of the [Training Model 28] with the highest accuracy.

[AI+ Analysis]
 - The AI+ analysis feature provides analysis and visualization of the data, aimed to help  users better understand statistical information.
 - Top 5 indices highly associated with a patient no-show are organized for users.

[Visualized information such as histograms and distribution charts are available for users.]

5. Predict Data | Make single or multiple predictions using your model

5.1 Single Prediction
 - Single Prediction allows you to directly enter your patient appointment information to predict  a customer no-show with that condition.
 - Predict patient no-shows for each individual patients by entering individual appointment information.

[Training Model 28 → Click on 'Single prediction']
 - Make single predictions using the [Training Model 28] with the highest accuracy.
 - 'Fill in random values' feature automatically fills all of the parameters randomly. After checking that all of the parameters have been filled appropriately, click 'Run' to proceed.

[The predicted result is displayed on the right.]
 - You may also manually enter values to make predictions.
 - If a value is out of range, the value will appear red.

5.2 Collective Prediction 
- Enter data to the downloaded collective prediction template, and run multiple no-show predictions.

[Training Model 28→ Click on 'Collective prediction']
 - Make collective predictions using the [Training Model 28] with the highest accuracy.

[Click on 'Download template for prediction' to download the prediction template.]

[Enter values to the prediction template (left), and prepare a prediction data set (right) to feed into the prediction model.]

[Upload the prepared prediction data, and proceed by pressing on 'Next'.]

[Prediction results will be available on a new csv file. The last column displays the prediction results for each patient.]
 - Collective prediction may require some time. You can download the prediction results from the notifications tab after completion.

5.3 Auto-Labeling
 - Enter data to the downloaded auto-labeling template, and run multiple no-show predictions.
 - Auto-labeling prediction will automatically preprocess data as needed.

[Training Model 28 → Click on 'Auto-labeling]
 - Make auto-labeling predictions using the [Training Model 28] with the highest accuracy.

[Click on 'Download template for prediction' to download the prediction template.]

[Enter values to the prediction template (left), and prepare a prediction data set (right) to feed into the prediction model.]

[Upload the prepared prediction data, and proceed by pressing on 'Next'.]

[Prediction results will be available on a new csv file. The last column displays the prediction results for each patient information.]
 - Auto-labeling prediction may require some time to complete. You can download the prediction results from the notifications tab after completion.

6. Share | Share your AI model with API

[Training Model 28→ Click on 'Details']

[Sharing your service app]
 - The URL link will direct you to shared page of the model.
 - Share your model using the URL link.
 - Others may have access to your model through the URL link. If somebody uses the model to make predictions, it will be counted from your prediction count.
 - 'Analyze' feature offers both explainable AI and visualizations on the shared link.
 - You can put your AI model on sale by using the AI store. For details, please check AI 스토어.

Consult with an expert now.

Consult with an expert now.