Upload a Dataset
This tutorial walks you through uploading a dataset to DashAI. Uploading a dataset is always the first step — once loaded, it becomes available across the platform for exploration, preprocessing, and model training.
Supported Formats
| Format | Dataloader | Extension |
|---|---|---|
| CSV | CSVDataLoader | .csv |
| Excel | ExcelDataLoader | .xlsx, .xls |
| JSON | JSONDataLoader | .json |
Step-by-Step Guide
1. Open the Datasets Section
In the top navigation bar, click on DATASETS. In the main area, click the "Upload Dataset" option at the center of the screen.
This opens the upload flow inline — you stay on the same screen throughout the entire process.
2. Select a Dataloader
A panel appears listing the available dataloaders, each with a brief description of the file format it handles.
Click the dataloader that matches your file (e.g., CSVDataLoader) and click "Next".
A Dataloader is the component that knows how to read a specific file format. Selecting the right one ensures DashAI can parse your file correctly and expose the appropriate configuration parameters for that format.
3. Upload Your File
Once a dataloader is selected, the file upload button becomes available. Click "Upload a File" and select your dataset file from your computer.
After the file is processed, DashAI displays a Dataset Preview in the center of the screen.
4. Review the Dataset Preview
The preview table shows the first 5 rows of a 100-row sample taken from your file. This sample is used by DashAI to automatically infer the data type of each column.
Showing 100 of 1000 rows analyzed for type inference.
You can change column types by clicking on the dropdown in each column header.
Editing column types
Each column header has a dropdown showing the inferred type. Click it to change it manually:
| Type | When to use |
|---|---|
Categorical | Discrete values representing groups or labels (e.g., gender, education level, home ownership status) |
Float | Continuous numbers with decimals (e.g., income, price, ratio) |
Integer | Whole numbers without decimals (e.g., age, count, years of experience) |
Text | Free-form natural language content (e.g., comments, descriptions, reviews) |
Reviewing and correcting column types at this stage ensures the platform interprets your data correctly in every subsequent step.
If you need to swap the file after seeing the preview, click the "RE-UPLOAD DATASET" button above the preview table. This takes you back to the file selection step without losing your current parameter configuration.
5. Configure the Parameters
The right panel is divided into two sections: Type Inference Configuration and Dataloader Configuration. Both affect how the dataset is read and stored.
Type Inference Configuration
This section appears for all dataloaders.
| Parameter | Type | Description |
|---|---|---|
| Inference Rows | Integer | The number of rows DashAI reads to automatically determine each column's data type. Default is 1000. Increasing this value improves inference accuracy on datasets with inconsistent or mixed values, at the cost of a slightly longer upload time. |
Dataloader Configuration
The parameters in this section depend on the dataloader you selected.
CSVDataLoader
| Parameter | Type | Description |
|---|---|---|
| Name | String | The name that will identify this dataset inside DashAI. Pre-filled from your filename — you can change it here. |
| Separator | Dropdown | The character that separates column values in your CSV. Default is , (comma). Switch to ; (semicolon) for files exported from Excel in Spanish or European locales, where comma is the decimal separator. |
| Header | String | How DashAI identifies the row containing column names. Default infer detects it automatically (usually the first row). Set a row number explicitly if your file has metadata lines before the actual header. |
| Names | String / Null | Optional list of column names to assign manually, overriding the names found in the file. Leave as Null to use the file's own column names. |
| Encoding | Dropdown | Character encoding of your file. Default is utf-8. Change to latin-1 or ISO-8859-1 if special characters (accents, symbols) appear garbled after upload. |
| NA values | String / Null | Additional strings to treat as missing values. For example, "?" or "N/A". Leave as Null to rely on default behavior. |
JSONDataLoader
| Parameter | Type | Description |
|---|---|---|
| Name | String | The name that will identify this dataset inside DashAI. Pre-filled from your filename — you can change it here. |
| Data key | String | The key inside your JSON file that contains the actual data records. Default is data. Change this to match the key in your file where the rows or records are located (e.g., "results", "records", "items"). |
DashAI expects your JSON file to contain a top-level object with a key pointing to
an array of records. For example: { "data": [ {...}, {...} ] }.
If your data is stored under a different key, set Data key accordingly.
ExcelDataLoader
| Parameter | Type | Description |
|---|---|---|
| Name | String | The name that will identify this dataset inside DashAI. Pre-filled from your filename — you can change it here. |
| Sheet | Integer | The index of the sheet to load, starting at 0. Default is 0 (the first sheet). Change this if your data is on a different sheet. |
| Header | Integer / Null | The row number (zero-indexed) that contains the column names. Default is 0 (first row). Set to Null if your file has no header row. |
| Use columns | String / Null | Specifies which columns to load. Leave as Null to load all columns. You can enter a comma-separated list of column names or indices to load only specific columns. |
| Skip rows | Integer / Null | Number of rows to skip at the start of the sheet before reading. Useful if your Excel file has title rows, report headers, or blank lines before the data. Leave as Null to skip nothing. |
| N rows | Integer / Null | Maximum number of rows to load. Leave as Null to load the entire sheet. Useful for testing with a large file. |
| Names | String / Null | Optional list of column names to assign manually. Leave as Null to use the column names found in the file. |
| NA values | String / Null | Additional strings to treat as missing values. Leave as Null to use default behavior. |
| Keep default NA | Checkbox | When enabled (default), DashAI recognizes a built-in list of common NA strings ("NA", "NaN", "null", empty strings, etc.) as missing values automatically. Disable only if you need full manual control. |
| True values | String / Null | Strings to interpret as boolean True (e.g., "yes", "Y", "1"). Leave as Null if your data has no boolean columns encoded as text. |
| False values | String / Null | Strings to interpret as boolean False (e.g., "no", "N", "0"). Leave as Null if not applicable. |
These two parameters work together. When Keep default NA is enabled and you also define custom NA values, both sets are combined — your custom strings are added on top of the defaults. Disabling Keep default NA means only your explicitly defined strings will be treated as missing.
6. Upload the Dataset
Once you have reviewed the column types and configured the parameters, click "UPLOAD" to finalize the process.
DashAI will process the file and add it to the Available Datasets list in the left sidebar, showing the total number of rows and columns.
Click "BACK" at any point to return to the previous step without losing your configuration.
Tips
- After uploading, use the Explorations module to visually verify that columns loaded correctly before running any experiment.
- If a numeric column appears as
Categorical, it likely contains non-numeric characters (spaces, symbols, or mixed values) — check the source file and re-upload after cleaning. - Use the N rows parameter (available in some dataloaders) to load only a portion of a large file while testing your configuration.
- The Separator parameter is the most common source of single-column load errors in CSV files. If your dataset appears as one long column, this is the first thing to check.
Troubleshooting
| Symptom | Likely cause | Solution |
|---|---|---|
| Dataset loads as a single column | Wrong separator selected (CSV) | Click RE-UPLOAD DATASET and set the correct separator |
A numeric column shows as Categorical | Column contains non-numeric characters | Fix the source file and re-upload |
| Special characters appear garbled | Wrong encoding (CSV) | Change Encoding to latin-1 or ISO-8859-1 |
| Column names are wrong or missing | Header row not at expected position | Adjust the Header parameter to the correct row index |
| Wrong sheet loaded (Excel) | Default sheet index is 0 | Change the Sheet parameter to the correct index |
| Extra rows at the top of the data (Excel) | File has title/metadata rows | Use Skip rows to skip them |
| NA values not recognized as missing | Custom NA strings not defined | Add your NA strings in the NA values field |
| JSON loads with wrong columns | Data is under a different key | Set Data key to the correct key in your JSON structure |