📊 Explore Data

Focus Question: What’s in my dataset, and how are my variables distributed, related, and missing?

1 Training Data

Training Data

(Completed Experiments Overview)

	input_1	input_2	input_3	input_4	input_5	input_6	input_7	input_8	output_1	output_2
0	100.0	DMAc	10.00	1.0	L29 | DPPF	0.015	(TMS)3SiH	0.6	9.0	77.0
1	100.0	DMSO	10.00	1.0	L29 | DPPF	0.015	(TMS)3SiH	0.6	0.0	91.0
2	100.0	NMP	10.00	1.0	L29 | DPPF	0.015	(TMS)3SiH	0.6	12.0	77.0
3	100.0	DMF	10.00	1.0	L29 | DPPF	0.015	(TMS)3SiH	0.6	2.0	78.0
4	100.0	DMPU	10.00	1.0	L29 | DPPF	0.015	(TMS)3SiH	0.6	6.0	91.0
5	100.0	Propionitrile	10.00	1.0	L29 | DPPF	0.015	(TMS)3SiH	0.6	0.0	98.0
6	100.0	DMAc	10.00	1.0	L33 | XantPhos	0.015	(TMS)3SiH	0.6	69.0	19.0
7	100.0	DMSO	10.00	1.0	L33 | XantPhos	0.015	(TMS)3SiH	0.6	0.0	86.0
8	100.0	NMP	10.00	1.0	L33 | XantPhos	0.015	(TMS)3SiH	0.6	54.0	37.0
9	100.0	DMF	10.00	1.0	L33 | XantPhos	0.015	(TMS)3SiH	0.6	6.0	87.0
10	100.0	DMPU	10.00	1.0	L33 | XantPhos	0.015	(TMS)3SiH	0.6	11.0	84.0
11	100.0	Propionitrile	10.00	1.0	L33 | XantPhos	0.015	(TMS)3SiH	0.6	0.0	94.0
12	100.0	DMAc	10.00	1.0	L59 | N-XantPhos	0.015	(TMS)3SiH	0.6	76.0	23.0
13	100.0	DMSO	10.00	1.0	L59 | N-XantPhos	0.015	(TMS)3SiH	0.6	0.0	91.0
14	100.0	NMP	10.00	1.0	L59 | N-XantPhos	0.015	(TMS)3SiH	0.6	0.0	100.0
15	100.0	DMF	10.00	1.0	L59 | N-XantPhos	0.015	(TMS)3SiH	0.6	0.0	90.0
16	100.0	DMPU	10.00	1.0	L59 | N-XantPhos	0.015	(TMS)3SiH	0.6	12.0	76.0
17	100.0	Propionitrile	10.00	1.0	L59 | N-XantPhos	0.015	(TMS)3SiH	0.6	0.0	96.0
18	100.0	DMAc	10.00	1.0	DPPP	0.015	(TMS)3SiH	0.6	89.0	2.0
19	100.0	DMSO	10.00	1.0	DPPP	0.015	(TMS)3SiH	0.6	1.0	91.0
20	100.0	NMP	10.00	1.0	DPPP	0.015	(TMS)3SiH	0.6	33.0	61.0
21	100.0	DMF	10.00	1.0	DPPP	0.015	(TMS)3SiH	0.6	32.0	62.0
22	100.0	DMPU	10.00	1.0	DPPP	0.015	(TMS)3SiH	0.6	38.0	51.0
23	100.0	Propionitrile	10.00	1.0	DPPP	0.015	(TMS)3SiH	0.6	0.0	90.0
24	100.0	DMAc	10.00	1.0	L29 | DPPF	0.015	PMHS	0.6	5.0	93.0
25	100.0	DMSO	10.00	1.0	L29 | DPPF	0.015	PMHS	0.6	4.0	94.0
26	100.0	NMP	10.00	1.0	L29 | DPPF	0.015	PMHS	0.6	6.0	85.0
27	100.0	DMF	10.00	1.0	L29 | DPPF	0.015	PMHS	0.6	7.0	80.0
28	100.0	DMPU	10.00	1.0	L29 | DPPF	0.015	PMHS	0.6	2.0	95.0
29	100.0	Propionitrile	10.00	1.0	L29 | DPPF	0.015	PMHS	0.6	15.0	88.0
30	100.0	DMAc	10.00	1.0	L33 | XantPhos	0.015	PMHS	0.6	5.0	89.0
31	100.0	DMSO	10.00	1.0	L33 | XantPhos	0.015	PMHS	0.6	18.0	71.0
32	100.0	NMP	10.00	1.0	L33 | XantPhos	0.015	PMHS	0.6	3.0	85.0
33	100.0	DMF	10.00	1.0	L33 | XantPhos	0.015	PMHS	0.6	8.0	79.0
34	100.0	DMPU	10.00	1.0	L33 | XantPhos	0.015	PMHS	0.6	3.0	95.0
35	100.0	Propionitrile	10.00	1.0	L33 | XantPhos	0.015	PMHS	0.6	4.0	90.0
36	100.0	DMAc	10.00	1.0	L59 | N-XantPhos	0.015	PMHS	0.6	0.0	91.0
37	100.0	DMSO	10.00	1.0	L59 | N-XantPhos	0.015	PMHS	0.6	5.0	92.0
38	100.0	NMP	10.00	1.0	L59 | N-XantPhos	0.015	PMHS	0.6	2.0	90.0
39	100.0	DMF	10.00	1.0	L59 | N-XantPhos	0.015	PMHS	0.6	2.0	88.0
40	100.0	DMPU	10.00	1.0	L59 | N-XantPhos	0.015	PMHS	0.6	0.0	99.0
41	100.0	Propionitrile	10.00	1.0	L59 | N-XantPhos	0.015	PMHS	0.6	1.0	93.0
42	100.0	DMAc	10.00	1.0	DPPP	0.015	PMHS	0.6	5.0	92.0
43	100.0	DMSO	10.00	1.0	DPPP	0.015	PMHS	0.6	10.0	82.0
44	100.0	NMP	10.00	1.0	DPPP	0.015	PMHS	0.6	3.0	86.0
45	100.0	DMF	10.00	1.0	DPPP	0.015	PMHS	0.6	5.0	88.0
46	100.0	DMPU	10.00	1.0	DPPP	0.015	PMHS	0.6	1.0	85.0
47	100.0	Propionitrile	10.00	1.0	DPPP	0.015	PMHS	0.6	7.0	85.0
48	110.0	DMAc	7.50	0.9	DPPP	0.020	(TMS)3SiH	0.3	72.0	13.0
49	120.0	DMAc	9.20	1.1	DPPP	0.036	(TMS)3SiH	0.6	86.0	0.0
50	110.0	DMAc	13.20	0.8	DPPP	0.037	(TMS)3SiH	0.1	66.0	20.0
51	90.0	DMAc	9.84	1.1	DPPP	0.037	(TMS)3SiH	0.6	76.0	21.0
52	100.0	DMAc	11.60	0.5	DPPP	0.024	(TMS)3SiH	0.2	60.0	30.0
53	110.0	DMAc	13.20	0.7	DPPP	0.041	(TMS)3SiH	0.2	89.0	0.0
54	90.0	DMAc	13.20	1.5	DPPP	0.040	PMHS	0.2	0.0	98.0
55	100.0	NMP	13.20	0.8	DPPP	0.042	(TMS)3SiH	0.2	87.0	7.0
56	110.0	DMAc	7.00	1.0	DPPP	0.015	(TMS)3SiH	0.1	84.0	0.0
57	110.0	DMAc	6.00	1.0	DPPP	0.015	(TMS)3SiH	0.1	85.0	0.0
58	110.0	DMAc	12.00	1.0	DPPP	0.015	(TMS)3SiH	0.1	87.0	0.0
59	100.0	DMAc	6.00	1.0	DPPP	0.037	(TMS)3SiH	0.1	82.0	14.0
60	100.0	DMAc	9.00	1.0	DPPP	0.030	(TMS)3SiH	0.1	86.0	0.0
61	90.0	DMAc	10.00	1.0	DPPP	0.015	(TMS)3SiH	0.1	88.0	0.0
62	120.0	NMP	13.00	1.0	DPPP	0.015	(TMS)3SiH	0.1	87.0	0.0
63	110.0	DMPU	10.00	1.0	DPPP	0.015	(TMS)3SiH	0.1	87.0	0.0

Table 1: Table of completed experiments in the training dataset.

2 Data Overview

Data Quality Warnings

(Warnings from training data)

Warning	Message
Correlation	Highly correlated pairs (|corr| > 0.90): Outputs 'output_1' and 'output_2': -0.99. Highly correlated inputs or outputs may reduce model robustness or optimization speed. Consider removing one of the variables if they carry redundant information.
Outliers	The following columns contain values that are unusually far from typical values: 'output_1': [66.0, 69.0, 72.0, 76.0, 82.0, 84.0, 85.0, 86.0, 87.0, 88.0, 89.0, 60.0] 'output_2': [0.0, 2.0, 7.0, 13.0, 14.0] Why does this matter? Outliers can affect the accuracy of machine learning models by skewing results or causing the model to focus too much on rare, extreme values. Consider reviewing these outliers to see if they represent errors in the data. If they are not errors, allow a few optimization rounds to better map the trends.

Table 2: Table summary of data quality warnings identified in the training dataset.

Summary of Numeric Columns

(Numeric variable summary statistics)

	input_1	input_3	input_4	input_6	input_8	output_1	output_2
count	64.000	64.000	64.000	64.000	64.000	64.000	64.000
mean	101.250	10.062	0.991	0.018	0.500	28.016	63.594
std	5.195	1.292	0.105	0.008	0.193	34.952	36.803
min	90.000	6.000	0.500	0.015	0.100	0.000	0.000
25%	100.000	10.000	1.000	0.015	0.600	2.000	22.500
50%	100.000	10.000	1.000	0.015	0.600	6.500	85.000
75%	100.000	10.000	1.000	0.015	0.600	66.750	91.000
max	120.000	13.200	1.500	0.042	0.600	89.000	100.000

Table 3: Table of summary statistics for numeric variables in the dataset.

Summary of Categorical Columns

(Categorical Variable Summary Statistics)

	input_2	input_5	input_7
count	64	64	64
unique	6	4	2
top	DMAc	DPPP	(TMS)3SiH
freq	21	28	39

Table 4: Table of summary statistics for categorical variables in the dataset.

3 Explore Variable Distributions

Variable Distributions

(Histogram matrix)

Histogram matrix

What it is: A grid of histograms showing how values for each variable are distributed.

How to use it:

• Each histogram shows the frequency of values for one variable.
• Look for skew, multimodality, or extreme outliers in the distributions.

Why it matters:

• Helps detect imbalances, skewness, or unusual ranges in variables.
• Provides an overview of the dataset’s coverage before modeling.

Limitations:

• Histograms can obscure fine-grained patterns in small datasets.
• Univariate distributions do not reveal relationships between variables.

Figure 1: Histogram matrix showing the distribution of all variables in the dataset.

Category–Output Comparisons

(Violin plots)

Violin plots

What it is: A visualization of how outputs vary across categories using violin plots.

How to use it:

• Each violin shows the distribution of an output for a category level.
• Wider sections indicate more frequent values; outliers may appear as separate points.

Why it matters:

• Reveals whether certain categories are associated with higher or lower outputs.
• Helps identify categories with more variability or uncertainty.

Limitations:

• Interpretation can be harder with very small category sample sizes.
• Does not capture interactions between multiple categorical variables.

(a) input_2-vs-output_1

(b) input_2-vs-output_2

(c) input_5-vs-output_1

(d) input_5-vs-output_2

(e) input_7-vs-output_1

(f) input_7-vs-output_2

Figure 2: Violin plots comparing category levels to output variables.

4 Variable Relationships

Variable Correlation

(Correlation heatmap)

Correlation heatmap

What it is: A heatmap summarizing linear relationships between numeric variables.

How to use it:

• Each cell shows a correlation coefficient (ranging from -1 to 1).
• Strong positive (near +1) or negative (near -1) values indicate strong relationships.
• Lighter colors near 0 indicate weak or no correlation.

Why it matters:

• Helps identify redundant variables that may not add information.
• Supports feature selection by showing which variables are strongly related.

Limitations:

• Captures only linear relationships, not non-linear associations.
• Correlation ≠ causation.
• Can obscure variable importance if not interpreted carefully.

Figure 3: Heatmap showing pairwise correlations between numeric variables.

Variable Association

(Mixed-type association heatmap)

Mixed-type association heatmap

What it is: A heatmap quantifying association strengths across continuous, categorical, and mixed variable types.

How to use it:

• Each cell shows association between two variables: squared correlation (continuous), Cramér’s V (categorical), η² (mixed).
• Values range from 0 (no association) to 1 (strong association).
• Look for strong blocks of association to detect redundancy.

Why it matters:

• Captures both linear and non-linear relationships across variable types.
• Guides feature selection and helps identify highly related inputs.

Limitations:

• Association ≠ causation.
• Dense heatmaps may be difficult to interpret without filtering.
• Can obscure variable importance if not interpreted carefully.

Figure 4: Heatmap showing associations between mixed variable types.