How should I structure my tabular data before uploading it to the platform?

Although the platform supports multiple file formats, the uploaded file must follow a structure in order for models to be successfully generated. Machine Learning tabular guidelines are not always compatible with typical engineering tabular practices. Ideally, the data should be uploaded as Tabular of 3D data. More details on file formats can be found here.

Requirements

When uploading tabular data to Monolith’s platform, please make sure it satisfies the following requirements.

General

• Each column should only contain one variable
• Rows should contain indices of experiment or sequence stamp (e.g. test number, simulation ID, time)
• Each column of the dataset should be classified as a type of data (e.g. string, float, etc.), thus it is not recommended to have more than one data type in the same column
• When using categorical data (i.e. strings), consistency is key. The platform assimilates “spaces” and is case sensitive, so “MonolithAI”, “Monolith AI” and “Monolith ai” are all classified as different categories
• The labelling of columns should only be 1 row (i.e. nested categorization should not be used in labelling)

Excel files

• The uploaded sheet should not have merged cells
• Font colour, font size and font type will not affect the data
• For Excel file: In case there are multiple sheets on the file being uploaded, only the first sheet will be uploaded
• Uploaded files should not contain images

Examples

The examples below highlight in red regions of the data that do not satisfy the requirements:

• Nested categorization of columns:

  	Front side			Rear side
  Time	Height	Force	Temp	Height	Force	Temp
  0.898260434	149.1872944	1610	62.50950145	102.9695565	1435	69.99225872
  0.06824658	126.9437422	1743	69.81139095	100.1677129	1719	72.39496438
  0.223245541	131.2395906	1704	70.92918715	104.4267423	1355	79.02104831
  0.198662073	138.8477895	1347	65.37057207	146.4874195	1380	62.5873467
  0.584823038	132.2786137	1406	70.06819137	131.7884798	1765	61.55247709

• Cells merged (and missing data):

  Time	F_Height	F_Force	F_Temp	R_Height	R_Force	R_Temp
  0.383839842	121.1055164	1688	63.7233866	101.4612084	1496
  0.484601597	125.1104664	1638		141.8651638	1495
  0.830386192	103.6675687	1773		136.6600568	1798	60.14558905
  0.93157364	122.4155221	1791		137.486305	1338	78.97866845
  0.067163228	105.3975624	1392	72.02537429	139.4831724	1501

• Inconsistent strings in a categorical column:

  Failure	Height	Force	Temp
  Compression	138.5704731	1362	76.61910893
  compression	117.1162992	1422	62.95551895
  Tension	108.2897529	1475	74.16961723
  in compression	104.5295069	1677	61.75505769
  tension	123.3339419	1770	78.61567623

• The example below is typical format of software outputs. However, for the platform each column must be a variable and rows must be a sequential stamp (e.g. time) or different experiments.

  Step 1
  Time	0.72615801
  Height	104.614631	Force:	1756	Temp:	72.38695876
  Step 2
  Time	0.452290895
  Height	139.1372833	Force:	1442	Temp:	74.92448148
  Step 3
  Time	0.624463615
  Height	147.5907217	Force:	1330	Temp:	64.89705344
  Step 4
  Time	0.323668304

• The example below follows the correct structure that data should be uploaded to the platform

  Time	F_Height	F_Force	F_Temp	R_Height	R_Force	R_Temp
  0.519042911	123.9004088	1567	71.39432069	120.7590165	1679	70.3345152
  0.902826988	124.4544405	1382	61.637524	121.0811838	1470	64.33987705
  0.869587281	109.1762983	1342	74.69825678	119.9179117	1707	62.49954938
  0.083198722	145.9851669	1407	60.32229119	144.0284513	1367	72.07843594
  0.18762859	122.550792	1584	65.540395	110.247237	1778	67.59115131