Application of Plastic Waste as a Sustainable Bitumen Mixture—A Review

Abstract

Plastic waste is growing rapidly, while asphalt binders remain heavily reliant on petroleum bitumen. Incorporating recycled plastics into bitumen can divert waste and enhance pavement performance. This review compiles 251 experimental records from 56 studies to evaluate how plastic type, dosage, and processing conditions affect softening point, penetration, and viscosity. Across studies, plastics (PET, LDPE/HDPE/LLDPE, PP, and hybrids) consistently stiffen binders, reducing penetration and increasing softening point and viscosity, thereby improving rutting resistance while potentially raising mixing/compaction demands. Using grouped cross-validated machine-learning models (median baseline, ridge, random forest, XGBoost), we quantify the predictability of binder properties and show that nonlinear methods outperform linear baselines for softening point. Prediction of penetration and viscosity shows larger scatter, reflecting study-to-study variability and incomplete reporting of key processing variables. We identify research needs in standardized testing, compatibility/dispersion characterization, and life-cycle assessment. The curated dataset and modeling workflow provide a data-driven foundation for designing durable, higher-performance plastic-modified binders.

1. Introduction

Plastic production has increased worldwide in recent decades, creating a substantial waste management problem. Landfilling and incineration are traditional methods that pose significant environmental risks, such as soil damage, greenhouse gas emissions, and ocean contamination. Since its inception, an estimated 8300 million tons of plastic have been produced [1]. By 2015, around 6300 million tons of garbage had been generated. Only 9% was recycled, 12% was burned, and the rest, 79%, wound up in landfills or the environment. If current trends continue, by 2050, approximately 12,000 million tons of plastic waste will have accumulated in landfills and natural places [1]. Natural resources are used extensively in road construction. Bitumen, the major binder in asphalt, is derived from petroleum; hence, its price and availability fluctuate with the market. There is currently a critical need to increase the quality of asphalt mixes while simultaneously making them more sustainable.

Thermoplastic polymers such as polyethylene (PE), polypropylene (PP), and polyethylene terephthalate (PET) contain long molecular chains that can enhance binder performance by improving viscoelastic behavior at service temperatures. When properly dispersed, these polymers increase stiffness and elasticity, resulting in a higher softening point, lower penetration, and greater viscosity, which together improve rutting resistance under heavy traffic. Plastics also exhibit good chemical compatibility with hydrocarbon-based binders, allowing them to act as functional modifiers rather than inert fillers. Compared with commercial virgin polymer modifiers, recycled plastics are low-cost, widely available, and help reduce landfill disposal, offering both performance improvements and sustainability benefits when used in asphalt mixtures.

The use of recycled plastic trash in bitumen has emerged as a viable road construction solution. This strategy not only reduces the quantity of plastic that ends up in landfills, but it also reduces reliance on virgin bitumen, which is expensive and restricted. By combining these advantages, the strategy promotes a more sustainable and circular economy, benefiting both waste management and the building industries [2]. Over the last 20 years, research has shown that plastics can operate as efficient modifiers, altering the rheological properties of bitumen to increase the performance of asphalt concrete in terms of rutting resistance, fatigue life, and durability. The findings revealed that the penetration values of the plastic-modified bitumen dropped as the amount of plastic increased. According to [3], the penetration values for 5%, 10%, and 20% plastic were 7.7 mm, 7.3 mm, and 2.0 mm, respectively. In contrast, as the plastic percentage increased, so did the softening point and ductility.

Another experiment was conducted where the doses of plastic used were 2, 4, 6, 8, and 10%, and it was concluded that replacing bitumen with plastic waste significantly improves the softening point, ductility, and penetration in flexible pavement construction, making it a cost-effective alternative to bitumen [4]. Also, with the same amount of dose, researcher [5] conducts tests on different quantities of 80/100 bitumen, which showed qualities such as elastic modulus, tensile strength, durability, resistance to permanent deformation, heat susceptibility, and fatigue resistance improved. An experiment conducted by [6], where he used 2, 3, 6, 8, and 10% of PET, showed a result that adding 3% PET waste improves the properties of modified bitumen, reducing penetration by 12 units and increasing the softening point by 5 °C.

This review paper compiles more than 251 data points from 56 published studies to provide a comprehensive understanding of plastic-modified bitumen. The analysis considers different types of plastics (PET, LDPE, HDPE, PP, hybrid, and PVC), dosage levels, mixing temperatures, and mixing rates, and evaluates their influence on key bitumen properties, including penetration, softening point, and viscosity. Machine learning models are applied to the combined dataset to improve predictive capability and uncover nonlinear relationships between processing conditions and performance outcomes. The review integrates existing knowledge, identifies consistent trends and research gaps, and proposes optimal modification parameters while addressing challenges and future opportunities for the wider adoption of plastic-modified bitumen in road construction.

Furthermore, this review incorporates a quantitative data-driven approach using machine learning to support the interpretation of the compiled experimental findings. Although this article remains a review study, a structured analytical methodology is employed to evaluate relationships between variables and enhance predictive understanding. This approach enables clearer synthesis of literature and strengthens the development of evidence-based recommendations for practical applications.

Positioning Relative to Recent Reviews

Several recent articles have surveyed the incorporation of waste and recycled plastics into bitumen, often focusing on experimental results, processing techniques, or qualitative trends. For example, the authors of [7] reviewed low-density and high-density PE modifiers; [8] focused on PET and PP waste plastics; and [9] addressed storage stability and phase separation in polymer-modified bitumen. However, most prior reviews suffer from one or more of the following limitations: they do not release the full dataset of experimental conditions, they do not evaluate predictability with cross-validated machine-learning models, and they provide limited or no coverage of hybrid/composite plastic systems (e.g., PP/PE blends, LDPE–LLDPE mixes) or sustainability aspects such as life-cycle assessment.

In contrast, the present study compiles 251 data points from 56 studies, standardizes key variables (polymer type, dose, and mixing parameters), trains grouped-cross-validated regression models (median, ridge, random forest, and XGBoost) to quantify the predictability of binder properties (softening point, penetration, and viscosity), reports averaged MAE, RMSE, and R² metrics, and publishes the complete dataset (as shown in Table 1). Furthermore, this review explicitly includes hybrid and composite plastic types and discusses environmental and economic implications. As such, we believe our work offers a novel dataset-driven analytical layer that complements and extends existing narrative reviews.

Table 1. Data set and reference.

Sl.no.	Type of Plastic Waste	Data			Reference
		Softening	Penetration	Viscosity
		(°C)	(dmm)	(Pa.S)
St_01	R-LLDPE	49.5	547	0.81	[10]
St_01	R-LLDPE	83.7	403	1.46	[10]
St_01	R-LLDPE	119.3	253	3.52	[10]
St_01	R-LLDPE	122.3	143	5.75	[10]
St_02	Plastic Waste	70	490	0.35	[11]
St_02	Plastic Waste	72	400	0.34	[11]
St_02	Plastic Waste	80	320	0.8	[11]
St_02	Plastic Waste	85	250	0.9	[11]
St_02	Plastic Waste	90	200	1.4	[11]
St_03	PET	43	550	NA	[12]
St_03	PET	51	450	NA	[12]
St_03	PET	62	400	NA	[12]
St_04	PET	47.5	820	6.5	[13]
St_04	PET	48.5	810	7.1	[13]
St_04	PET	50	780	7.55	[13]
St_04	PET	51.25	740	7.8	[13]
St_04	PET	53	710	7.9	[13]
St_04	PET	55	670	8.1	[13]
St_05	PET	NA	NA	NA	[14]
St_05	PET	NA	NA	NA	[14]
St_05	PET	NA	NA	NA	[14]
St_05	PET	NA	NA	NA	[14]
St_06	Hybrid	80	250	NA	[15]
St_06	Hybrid	75	250	NA	[15]
St_06	Hybrid	90	240	NA	[15]
St_07	LDPE	43	730	NA	[4]
St_07	LDPE	48	580	NA	[4]
St_07	LDPE	57	550	NA	[4]
St_07	LDPE	61	530	NA	[4]
St_07	LDPE	63	500	NA	[4]
St_07	LDPE	66	460	NA	[4]
St_08	LDPE	NA	420	NA	[14]
St_08	LDPE	NA	400	NA	[14]
St_08	LDPE	NA	400	NA	[14]
St_08	LDPE	NA	420	NA	[14]
St_09	PET	NA	NA	NA	[16]
St_09	HDPE	NA	NA	NA	[16]
St_09	PP	NA	NA	NA	[16]
St_09	PS	NA	NA	NA	[16]
St_09	PVC	NA	NA	NA	[16]
St_10	HP	NA	NA	2.4	[17]
St_10	HP	NA	NA	0.9	[17]
St_10	HP	NA	NA	0.5	[17]
St_10	HP	NA	NA	1.2	[17]
St_11	PVC	42	97	NA	[18]
St_11	PVC	43	91	NA	[18]
St_11	PVC	43.55	84	NA	[18]
St_12	Hybrid	67.33	29	NA	[5]
St_12	Hybrid	83.33	59	NA	[5]
St_12	Hybrid	73	56	NA	[5]
St_12	Hybrid	73	53	NA	[5]
St_12	Hybrid	115	47	NA	[5]
St_13	PVC	40.5	165	0.245	[18]
St_13	PVC	41.3	157	0.282	[18]
St_13	PVC	42	150	0.298	[18]
St_14	PP	55	760	NA	[19]
St_14	PP	59	710	NA	[19]
St_14	PP	61	660	NA	[19]
St_14	PP	66	590	NA	[19]
St_14	PP	71	530	NA	[19]
St_14	PP	78	490	NA	[19]
St_14	PP	81	450	NA	[19]
St_14	PP	83	390	NA	[19]
St_15	PVC	6	463	NA	[18]
St_15	PVC	5.3	490	NA	[18]
St_15	PVC	3.833	540	NA	[18]
St_16	R-LLDPE	44.1	593	0.62	[10]
St_16	R-LLDPE	49.5	547	0.81	[10]
St_16	R-LLDPE	83.7	403	1.46	[10]
St_16	R-LLDPE	119.3	253	3.52	[10]
St_16	R-LLDPE	122.3	143	5.57	[10]
St_17	PVC	45	81	NA	[18]
St_17	PVC	46.25	73	NA	[18]
St_17	PVC	51	83	NA	[18]
St_18	LDPE and LLDPE	10	NA	NA	[20]
St_18	LDPE and LLDPE	5	NA	NA	[20]
St_18	LDPE and LLDPE	29	NA	NA	[20]
St_18	LDPE and LLDPE	130	NA	NA	[20]
St_18	LDPE and LLDPE	131	NA	NA	[20]
St_19	PVC	41.05	135	0.25	[18]
St_19	PVC	41.85	122	0.3	[18]
St_19	PVC	42.25	106	0.39	[18]
St_20	PET	NA	450	NA	[21]
St_20	PET	NA	420	NA	[21]
St_20	PET	NA	390	NA	[21]
St_20	PET	NA	350	NA	[21]
St_20	PET	NA	300	NA	[21]
St_20	HDPE	NA	450	NA	[21]
St_20	HDPE	NA	440	NA	[21]
St_20	HDPE	NA	430	NA	[21]
St_20	HDPE	NA	460	NA	[21]
St_20	HDPE	NA	470	NA	[21]
St_20	LDPE	NA	460	NA	[21]
St_20	LDPE	NA	450	NA	[21]
St_20	LDPE	NA	410	NA	[21]
St_20	LDPE	NA	415	NA	[21]
St_20	LDPE	NA	450	NA	[21]
St_21	PET	46.9	80	0.38	[22]
St_21	PET	48.4	73	0.44	[22]
St_21	PET	48.8	65	0.41	[22]
St_21	PET	50.1	60	0.45	[22]
St_22	PET	NA	101.5	0.65	[5]
St_22	PET	44.5	102.75	0.73	[5]
St_22	PET	44.8	103.5	0.79	[5]
St_22	PET	45.2	104.75	0.86	[5]
St_22	PET	47	110	0.91	[5]
St_22	PET	47.5	112	0.95	[5]
St_23	PP	55.7	47	NA	[23]
St_23	PP	54.4	49	NA	[23]
St_23	PP	53.9	53	NA	[23]
St_24	HDPE	NA	NA	NA	[24]
St_24	HDPE	NA	NA	NA	[24]
St_24	HDPE	NA	NA	NA	[24]
St_24	HDPE	NA	NA	NA	[24]
St_24	LDPE	NA	NA	NA	[24]
St_24	LDPE	NA	NA	NA	[24]
St_24	LDPE	NA	NA	NA	[24]
St_25	PET	52	730	NA	[6]
St_25	PET	54	650	NA	[6]
St_25	PET	57	610	NA	[6]
St_25	PET	56	620	NA	[6]
St_25	PET	55	640	NA	[6]
St_26	R-LLDPE	50	55	0.8	[10]
St_26	R-LLDPE	80	40	1.5	[10]
St_26	R-LLDPE	118	25	3.5	[10]
St_26	R-LLDPE	120	15	6.7	[10]
St_27	HDPE	62.2	323	0.54	[25]
St_27	HDPE	68.6	285	0.78	[25]
St_27	HDPE	73.5	218	1.28	[25]
St_27	HDPE	75.7	199	1.42	[25]
St_27	HDPE	74	213	1.54	[25]
St_28	PET	56	422	NA	[26]
St_28	PET	59.4	400	NA	[26]
St_28	PET	60.1	383	NA	[26]
St_28	PET	61	361	NA	[26]
St_29	PE-P	66	NA	1.5	[9]
St_29	PE-S	90	NA	2.7	[9]
St_30	PP/PE	65.1	470	0.618	[8]
St_30	PP/PE	61.9	473	0.609	[8]
St_30	PP/PE	63.6	464	0.626	[8]
St_30	PP/PE	63.3	475	0.606	[8]
St_30	PP/PE	62.9	473	0.612	[8]
St_30	PP/PE	62.6	484	0.6	[8]
St_30	PP/PE	62.5	473	0.617	[8]
St_31	PET	45	666	0.196	[7]
St_31	PET	46.5	649	0.238	[7]
St_31	PET	48	638	0.293	[7]
St_31	PET	50.5	612	0.341	[7]
St_31	PET	53	594	0.407	[7]
St_31	PET	55.5	573	0.459	[7]
St_31	PET	56	569	0.537	[7]
St_32	LDPE	52.5	55	0.54	[27]
St_32	LDPE	52.7	49	0.72	[27]
St_32	LDPE	61.8	38	0.84	[27]
St_33	SBS	NA	NA	NA	[28]
St_33	SBS	NA	NA	NA	[28]
St_33	SBS	NA	NA	NA	[28]
St_33	EVA	NA	NA	NA	[28]
St_33	EVA	NA	NA	NA	[28]
St_33	EVA	NA	NA	NA	[28]
St_33	MR6	NA	NA	NA	[28]
St_33	MR10	NA	NA	NA	[28]
St_34	Waste Plastic	45	55	NA	[12]
St_34	Waste Plastic	47.5	53	NA	[12]
St_34	Waste Plastic	53	43	NA	[12]
St_35	HDPE	NA	NA	NA	[29]
St_35	HDPE	NA	NA	NA	[29]
St_35	HDPE	NA	NA	NA	[29]
St_36	VPP	51.6	45	0.2	[30]
St_36	VPP	50.3	50	0.18	[30]
St_36	VPP	50.1	54	0.175	[30]
St_36	VPP	50	57	0.17	[30]
St_36	VPP	49.8	60	0.16	[30]
St_37	LDPE	63	320	0.7	[31]
St_37	HDPE	61	270	0.578	[31]
St_37	LLDPE	59	210	0.38	[31]
St_37	PP	63	240	0.6	[31]
St_37	EVA	57	500	0.98	[31]
St_37	EBA	58	560	0.94	[31]
St_38	PET	75	730	NA	[32]
St_38	PET	78	710	NA	[32]
St_38	PET	85	670	NA	[32]
St_38	PET	98	560	NA	[32]
St_38	PET	99	530	NA	[32]
St_38	PET	100	500	NA	[32]
St_39	LDPE	77	300	NA	[33]
St_39	LDPE	77	300	NA	[33]
St_39	LDPE	77	300	NA	[33]
St_39	LDPE	77	300	NA	[33]
St_39	HDPE	87	255	NA	[33]
St_39	HDPE	87	255	NA	[33]
St_39	HDPE	87	255	NA	[33]
St_39	HDPE	87	255	NA	[33]
St_40	PET	47	650	NA	[34]
St_40	PET	47.5	620	NA	[34]
St_40	PET	48.5	600	NA	[34]
St_41	PVC	80.7	670	NA	[35]
St_41	PVC	81.2	490	NA	[35]
St_42	Waste Plastic	50	138	2	[36]
St_42	Waste Plastic	55	100	4	[36]
St_42	Waste Plastic	54	90	4.1	[36]
St_42	Waste Plastic	54	85	4.9	[36]
St_42	Waste Plastic	55	85	3.5	[36]
St_42	Waste Plastic	59	80	5.8	[36]
St_43	PVC	55	550	4.3	[37]
St_43	PVC	58	440	5	[37]
St_44	Hybrid	NA	NA	NA	[38]
St_44	Hybrid	NA	NA	NA	[38]
St_44	Hybrid	NA	NA	NA	[38]
St_44	Hybrid	NA	NA	NA	[38]
St_45	PET	52	49	NA	[39]
St_45	PET	53	35	NA	[39]
St_45	PET	57	29	NA	[39]
St_46	PET	46.5	649	3.73	[40]
St_46	PET	48	638	3.93	[40]
St_46	PET	50.5	612	4.02	[40]
St_46	PET	53	594	4.13	[40]
St_46	PET	55.5	573	4.59	[40]
St_46	PET	56	569	4.88	[40]
St_46	PET	56.7	550	4.93	[40]
St_47	E-waste plastic	54	643	0.99	[41]
St_47	E-waste plastic	55	114.33	1.14	[41]
St_47	E-waste plastic	66.7	123.33	1.46	[41]
St_47	E-waste plastic	74.5	113.33	1.86	[41]
St_47	E-waste plastic	85	313	2.01	[41]
St_48	Plastic	48.5	70	NA	[42]
St_48	Plastic	45	59	NA	[42]
St_48	Crumbed rubber	47.5	68	NA	[42]
St_48	Crumbed rubber	42	54	NA	[42]
St_49	PET+TETA	NA	NA	5.3	[43]
St_49	PET+EA	NA	NA	5.05	[43]
St_50	HDPE	54	140	NA	[36]
St_50	HDPE	56	89	NA	[36]
St_50	HDPE	57	90	NA	[36]
St_50	HDPE	58	81	NA	[36]
St_50	HDPE	61	80	NA	[36]
St_50	HDPE	64	75	NA	[36]
St_50	HDPE	70	70	NA	[36]
St_50	PP	54	140	NA	[36]
St_50	PP	50	138	NA	[36]
St_50	PP	55	103	NA	[36]
St_50	PP	53	90	NA	[36]
St_50	PP	54	86	NA	[36]
St_50	PP	53	85	NA	[36]
St_50	PP	59	80	NA	[36]
St_51	PET	45	75	4	[44]
St_51	PET	46.5	50	4.5	[44]
St_51	PET	47	40	6	[44]
St_51	PET	51	38	8.5	[44]
St_51	PET	52	25	10	[44]

Footnote: Some entries exhibit apparent outliers (e.g., penetration values above 100 dmm) due to study-specific reporting scales or binder grades. All values were preserved as reported and standardized during modeling. Missing values (“NA”) were retained in the curated dataset to avoid imputation bias and were excluded only for the specific target variable in training/evaluation (GroupKFold by Study_ID). See Data Dictionary (Table 2) for null counts and completeness, and Table 1 for full records.

Table 1. Data set and reference.

Sl.no.	Type of Plastic Waste	Data			Reference
		Softening	Penetration	Viscosity
		(°C)	(dmm)	(Pa.S)
St_01	R-LLDPE	49.5	547	0.81	[10]
St_01	R-LLDPE	83.7	403	1.46	[10]
St_01	R-LLDPE	119.3	253	3.52	[10]
St_01	R-LLDPE	122.3	143	5.75	[10]
St_02	Plastic Waste	70	490	0.35	[11]
St_02	Plastic Waste	72	400	0.34	[11]
St_02	Plastic Waste	80	320	0.8	[11]
St_02	Plastic Waste	85	250	0.9	[11]
St_02	Plastic Waste	90	200	1.4	[11]
St_03	PET	43	550	NA	[12]
St_03	PET	51	450	NA	[12]
St_03	PET	62	400	NA	[12]
St_04	PET	47.5	820	6.5	[13]
St_04	PET	48.5	810	7.1	[13]
St_04	PET	50	780	7.55	[13]
St_04	PET	51.25	740	7.8	[13]
St_04	PET	53	710	7.9	[13]
St_04	PET	55	670	8.1	[13]
St_05	PET	NA	NA	NA	[14]
St_05	PET	NA	NA	NA	[14]
St_05	PET	NA	NA	NA	[14]
St_05	PET	NA	NA	NA	[14]
St_06	Hybrid	80	250	NA	[15]
St_06	Hybrid	75	250	NA	[15]
St_06	Hybrid	90	240	NA	[15]
St_07	LDPE	43	730	NA	[4]
St_07	LDPE	48	580	NA	[4]
St_07	LDPE	57	550	NA	[4]
St_07	LDPE	61	530	NA	[4]
St_07	LDPE	63	500	NA	[4]
St_07	LDPE	66	460	NA	[4]
St_08	LDPE	NA	420	NA	[14]
St_08	LDPE	NA	400	NA	[14]
St_08	LDPE	NA	400	NA	[14]
St_08	LDPE	NA	420	NA	[14]
St_09	PET	NA	NA	NA	[16]
St_09	HDPE	NA	NA	NA	[16]
St_09	PP	NA	NA	NA	[16]
St_09	PS	NA	NA	NA	[16]
St_09	PVC	NA	NA	NA	[16]
St_10	HP	NA	NA	2.4	[17]
St_10	HP	NA	NA	0.9	[17]
St_10	HP	NA	NA	0.5	[17]
St_10	HP	NA	NA	1.2	[17]
St_11	PVC	42	97	NA	[18]
St_11	PVC	43	91	NA	[18]
St_11	PVC	43.55	84	NA	[18]
St_12	Hybrid	67.33	29	NA	[5]
St_12	Hybrid	83.33	59	NA	[5]
St_12	Hybrid	73	56	NA	[5]
St_12	Hybrid	73	53	NA	[5]
St_12	Hybrid	115	47	NA	[5]
St_13	PVC	40.5	165	0.245	[18]
St_13	PVC	41.3	157	0.282	[18]
St_13	PVC	42	150	0.298	[18]
St_14	PP	55	760	NA	[19]
St_14	PP	59	710	NA	[19]
St_14	PP	61	660	NA	[19]
St_14	PP	66	590	NA	[19]
St_14	PP	71	530	NA	[19]
St_14	PP	78	490	NA	[19]
St_14	PP	81	450	NA	[19]
St_14	PP	83	390	NA	[19]
St_15	PVC	6	463	NA	[18]
St_15	PVC	5.3	490	NA	[18]
St_15	PVC	3.833	540	NA	[18]
St_16	R-LLDPE	44.1	593	0.62	[10]
St_16	R-LLDPE	49.5	547	0.81	[10]
St_16	R-LLDPE	83.7	403	1.46	[10]
St_16	R-LLDPE	119.3	253	3.52	[10]
St_16	R-LLDPE	122.3	143	5.57	[10]
St_17	PVC	45	81	NA	[18]
St_17	PVC	46.25	73	NA	[18]
St_17	PVC	51	83	NA	[18]
St_18	LDPE and LLDPE	10	NA	NA	[20]
St_18	LDPE and LLDPE	5	NA	NA	[20]
St_18	LDPE and LLDPE	29	NA	NA	[20]
St_18	LDPE and LLDPE	130	NA	NA	[20]
St_18	LDPE and LLDPE	131	NA	NA	[20]
St_19	PVC	41.05	135	0.25	[18]
St_19	PVC	41.85	122	0.3	[18]
St_19	PVC	42.25	106	0.39	[18]
St_20	PET	NA	450	NA	[21]
St_20	PET	NA	420	NA	[21]
St_20	PET	NA	390	NA	[21]
St_20	PET	NA	350	NA	[21]
St_20	PET	NA	300	NA	[21]
St_20	HDPE	NA	450	NA	[21]
St_20	HDPE	NA	440	NA	[21]
St_20	HDPE	NA	430	NA	[21]
St_20	HDPE	NA	460	NA	[21]
St_20	HDPE	NA	470	NA	[21]
St_20	LDPE	NA	460	NA	[21]
St_20	LDPE	NA	450	NA	[21]
St_20	LDPE	NA	410	NA	[21]
St_20	LDPE	NA	415	NA	[21]
St_20	LDPE	NA	450	NA	[21]
St_21	PET	46.9	80	0.38	[22]
St_21	PET	48.4	73	0.44	[22]
St_21	PET	48.8	65	0.41	[22]
St_21	PET	50.1	60	0.45	[22]
St_22	PET	NA	101.5	0.65	[5]
St_22	PET	44.5	102.75	0.73	[5]
St_22	PET	44.8	103.5	0.79	[5]
St_22	PET	45.2	104.75	0.86	[5]
St_22	PET	47	110	0.91	[5]
St_22	PET	47.5	112	0.95	[5]
St_23	PP	55.7	47	NA	[23]
St_23	PP	54.4	49	NA	[23]
St_23	PP	53.9	53	NA	[23]
St_24	HDPE	NA	NA	NA	[24]
St_24	HDPE	NA	NA	NA	[24]
St_24	HDPE	NA	NA	NA	[24]
St_24	HDPE	NA	NA	NA	[24]
St_24	LDPE	NA	NA	NA	[24]
St_24	LDPE	NA	NA	NA	[24]
St_24	LDPE	NA	NA	NA	[24]
St_25	PET	52	730	NA	[6]
St_25	PET	54	650	NA	[6]
St_25	PET	57	610	NA	[6]
St_25	PET	56	620	NA	[6]
St_25	PET	55	640	NA	[6]
St_26	R-LLDPE	50	55	0.8	[10]
St_26	R-LLDPE	80	40	1.5	[10]
St_26	R-LLDPE	118	25	3.5	[10]
St_26	R-LLDPE	120	15	6.7	[10]
St_27	HDPE	62.2	323	0.54	[25]
St_27	HDPE	68.6	285	0.78	[25]
St_27	HDPE	73.5	218	1.28	[25]
St_27	HDPE	75.7	199	1.42	[25]
St_27	HDPE	74	213	1.54	[25]
St_28	PET	56	422	NA	[26]
St_28	PET	59.4	400	NA	[26]
St_28	PET	60.1	383	NA	[26]
St_28	PET	61	361	NA	[26]
St_29	PE-P	66	NA	1.5	[9]
St_29	PE-S	90	NA	2.7	[9]
St_30	PP/PE	65.1	470	0.618	[8]
St_30	PP/PE	61.9	473	0.609	[8]
St_30	PP/PE	63.6	464	0.626	[8]
St_30	PP/PE	63.3	475	0.606	[8]
St_30	PP/PE	62.9	473	0.612	[8]
St_30	PP/PE	62.6	484	0.6	[8]
St_30	PP/PE	62.5	473	0.617	[8]
St_31	PET	45	666	0.196	[7]
St_31	PET	46.5	649	0.238	[7]
St_31	PET	48	638	0.293	[7]
St_31	PET	50.5	612	0.341	[7]
St_31	PET	53	594	0.407	[7]
St_31	PET	55.5	573	0.459	[7]
St_31	PET	56	569	0.537	[7]
St_32	LDPE	52.5	55	0.54	[27]
St_32	LDPE	52.7	49	0.72	[27]
St_32	LDPE	61.8	38	0.84	[27]
St_33	SBS	NA	NA	NA	[28]
St_33	SBS	NA	NA	NA	[28]
St_33	SBS	NA	NA	NA	[28]
St_33	EVA	NA	NA	NA	[28]
St_33	EVA	NA	NA	NA	[28]
St_33	EVA	NA	NA	NA	[28]
St_33	MR6	NA	NA	NA	[28]
St_33	MR10	NA	NA	NA	[28]
St_34	Waste Plastic	45	55	NA	[12]
St_34	Waste Plastic	47.5	53	NA	[12]
St_34	Waste Plastic	53	43	NA	[12]
St_35	HDPE	NA	NA	NA	[29]
St_35	HDPE	NA	NA	NA	[29]
St_35	HDPE	NA	NA	NA	[29]
St_36	VPP	51.6	45	0.2	[30]
St_36	VPP	50.3	50	0.18	[30]
St_36	VPP	50.1	54	0.175	[30]
St_36	VPP	50	57	0.17	[30]
St_36	VPP	49.8	60	0.16	[30]
St_37	LDPE	63	320	0.7	[31]
St_37	HDPE	61	270	0.578	[31]
St_37	LLDPE	59	210	0.38	[31]
St_37	PP	63	240	0.6	[31]
St_37	EVA	57	500	0.98	[31]
St_37	EBA	58	560	0.94	[31]
St_38	PET	75	730	NA	[32]
St_38	PET	78	710	NA	[32]
St_38	PET	85	670	NA	[32]
St_38	PET	98	560	NA	[32]
St_38	PET	99	530	NA	[32]
St_38	PET	100	500	NA	[32]
St_39	LDPE	77	300	NA	[33]
St_39	LDPE	77	300	NA	[33]
St_39	LDPE	77	300	NA	[33]
St_39	LDPE	77	300	NA	[33]
St_39	HDPE	87	255	NA	[33]
St_39	HDPE	87	255	NA	[33]
St_39	HDPE	87	255	NA	[33]
St_39	HDPE	87	255	NA	[33]
St_40	PET	47	650	NA	[34]
St_40	PET	47.5	620	NA	[34]
St_40	PET	48.5	600	NA	[34]
St_41	PVC	80.7	670	NA	[35]
St_41	PVC	81.2	490	NA	[35]
St_42	Waste Plastic	50	138	2	[36]
St_42	Waste Plastic	55	100	4	[36]
St_42	Waste Plastic	54	90	4.1	[36]
St_42	Waste Plastic	54	85	4.9	[36]
St_42	Waste Plastic	55	85	3.5	[36]
St_42	Waste Plastic	59	80	5.8	[36]
St_43	PVC	55	550	4.3	[37]
St_43	PVC	58	440	5	[37]
St_44	Hybrid	NA	NA	NA	[38]
St_44	Hybrid	NA	NA	NA	[38]
St_44	Hybrid	NA	NA	NA	[38]
St_44	Hybrid	NA	NA	NA	[38]
St_45	PET	52	49	NA	[39]
St_45	PET	53	35	NA	[39]
St_45	PET	57	29	NA	[39]
St_46	PET	46.5	649	3.73	[40]
St_46	PET	48	638	3.93	[40]
St_46	PET	50.5	612	4.02	[40]
St_46	PET	53	594	4.13	[40]
St_46	PET	55.5	573	4.59	[40]
St_46	PET	56	569	4.88	[40]
St_46	PET	56.7	550	4.93	[40]
St_47	E-waste plastic	54	643	0.99	[41]
St_47	E-waste plastic	55	114.33	1.14	[41]
St_47	E-waste plastic	66.7	123.33	1.46	[41]
St_47	E-waste plastic	74.5	113.33	1.86	[41]
St_47	E-waste plastic	85	313	2.01	[41]
St_48	Plastic	48.5	70	NA	[42]
St_48	Plastic	45	59	NA	[42]
St_48	Crumbed rubber	47.5	68	NA	[42]
St_48	Crumbed rubber	42	54	NA	[42]
St_49	PET+TETA	NA	NA	5.3	[43]
St_49	PET+EA	NA	NA	5.05	[43]
St_50	HDPE	54	140	NA	[36]
St_50	HDPE	56	89	NA	[36]
St_50	HDPE	57	90	NA	[36]
St_50	HDPE	58	81	NA	[36]
St_50	HDPE	61	80	NA	[36]
St_50	HDPE	64	75	NA	[36]
St_50	HDPE	70	70	NA	[36]
St_50	PP	54	140	NA	[36]
St_50	PP	50	138	NA	[36]
St_50	PP	55	103	NA	[36]
St_50	PP	53	90	NA	[36]
St_50	PP	54	86	NA	[36]
St_50	PP	53	85	NA	[36]
St_50	PP	59	80	NA	[36]
St_51	PET	45	75	4	[44]
St_51	PET	46.5	50	4.5	[44]
St_51	PET	47	40	6	[44]
St_51	PET	51	38	8.5	[44]
St_51	PET	52	25	10	[44]

Footnote: Some entries exhibit apparent outliers (e.g., penetration values above 100 dmm) due to study-specific reporting scales or binder grades. All values were preserved as reported and standardized during modeling. Missing values (“NA”) were retained in the curated dataset to avoid imputation bias and were excluded only for the specific target variable in training/evaluation (GroupKFold by Study_ID). See Data Dictionary (Table 2) for null counts and completeness, and Table 1 for full records.

Table 2. Hyperparameter tuning grids.

Model	Key Hyperparameters	Search Range
XGBoost	Learning rate; max. depth; n_estimators; subsample; colsample_bytree; reg_lambda	0.01–0.3; 3–10; 100–500; 0.6–1.0; 0.6–1.0; 0–2.0
Rando Forest	n_estimators; max. depth; min_samples_split; min_samples_leaf; max_features	200–800, 3–20, 2–10, 1–5 (auto square)
Ridge Regression	alpha (L2 regularization)	10(−04)–10(04) (log-spaced)

Table 2. Hyperparameter tuning grids.

Model	Key Hyperparameters	Search Range
XGBoost	Learning rate; max. depth; n_estimators; subsample; colsample_bytree; reg_lambda	0.01–0.3; 3–10; 100–500; 0.6–1.0; 0.6–1.0; 0–2.0
Rando Forest	n_estimators; max. depth; min_samples_split; min_samples_leaf; max_features	200–800, 3–20, 2–10, 1–5 (auto square)
Ridge Regression	alpha (L2 regularization)	10(−04)–10(04) (log-spaced)

2. Methodology for Data Extraction and Analysis

A dataset of 251 experimental records was compiled from 56 published studies, covering three commonly reported binder properties: softening point (°C), penetration (d.mm), and viscosity (Pa.s). Each row corresponds to one unique experimental condition and includes details on polymer type, dosage (%), mixing temperature (°C), mixing rate (rpm), mixing time (min), bitumen grade, and aging condition. A Study_ID field was added to retain the source reference for reproducibility and validation control.

Categorical fields (e.g., plastic type, bitumen grade) were encoded as categories, and numeric fields were stored as floats. Where source studies reported different unit scales, the reported values were preserved and standardized for modeling. Missing values (NA) were retained to avoid introducing bias and excluded only for the specific target variable during training. To assess generalization across studies, all model evaluations used a five-fold GroupKFold, where Study_ID was used as the grouping key. This ensured that data from the same study never appeared simultaneously in both training and validation folds.

Machine Learning Workflow

Three regression tasks were modeled independently: softening point, penetration, and viscosity to preserve interpretability for each binder property. Four models were evaluated:

• Median baseline;
• Ridge regression;
• Random forest;
• XGBoost regression.

All models were implemented in Python (version 3.12.12) using scikit-learn and XGBoost. Categorical predictors were one-hot encoded; numeric features remained on their native scales for tree-based models, while ridge employed internal scaling.

Model hyperparameters were tuned within folds using GroupKFold (Study_ID) with grid/random search. Example grids are as follows:

Hyperparameters were optimized using grid/random search within each training fold, and performance was averaged across folds using the following:

• Mean Absolute Error (MAE);
• Root Mean Square Error (RMSE);
• Coefficient of Determination (R²).

Feature contribution was analyzed using permutation importance and SHAP value inspection to evaluate sensitivity to key variables such as polymer type and dosage.

Mean absolute error (MAE)

It tells how far a model’s predictions are from the actual values on average, without regard for whether they were too high or too low. A high MAE could indicate that the model should be retrained or modified. A lower MAE increases confidence that the model generates useful estimates [45]. According to the research [46], the equation to find the mean absolute error is

$$\mathrm{MAE}={\displaystyle \frac{1}{n}}{\sum }_{i=1}^n(Y_i+Y_i^{\prime }{)}^2$$

Root mean square error (RMSE)

The difference between predicted and actual values in a regression model is measured, and the standard deviation of the residuals (errors) is used to determine its accuracy. A lower RMSE suggests that the model fits the data better, with 0 indicating a perfect model [47]. The root mean square error is determined as follows:

$$\mathrm{RMSE} = \sqrt{{\displaystyle \frac{1}{n}}{\sum }_{i=1}^n(Y_i+Y_i^{\prime }{)}^2}$$

Coefficient of Determination (R²)

The coefficient of determination is a summary statistic that shows how well the independent variable in the regression explains the variation in the dependent variable [48]. The coefficient of determination R² is the following ratio:

$$R^2={\displaystyle \frac{Explained Variation}{Total Variation}}={\displaystyle \frac{RSS}{TSS}}$$

$$\mathrm{RMSE}=1-{\displaystyle \frac{\sum (Y_i+Y_i^{\prime }{)}^2}{\sum (Y_i+Y_i^{\prime }{)}^2}}$$

To supplement these measurements, diagnostic charts such as predicted vs. observed charts and feature importance plots were created.

Table 3. Data dictionary.

Column Name	Data Type	Data Category	Sample Value	Null Count	Completeness (%)	Unique Value	Value Range	Mean
Study_id	object	Text	St_01, St_01, St_01	0	100	51	NA	NA
Type of plastic	object	Text	R-LLDPE, R-LLDPE, R-LLDPE	0	100	28	NA	NA
Dose of plastic (%)	float64	Numeric	3.0, 3.0, 3.0	0	100	25	0–36	5.43
Mixing Temp.	float64	Numeric	170.0, 170.0, 170.0	15	94	17	120–250	170
Mixing Rate	float64	Numeric	3500.0, 3500.0, 3500.0	85	66	16	20–13,000	2110
Mixing Time	float64	Numeric	90.0, 90.0, 90.0	42	83	17	2–180	49
Type of Bitumen	object	Text	C320, C320, C320	11	95	39	NA	NA
Age/unage	object	Text	Unaged, Unaged, Unaged	231	7.6	2	NA	NA
Softening point	float64	Numeric	49.5, 83.7, 119.3	57	77.2	102	3.83–131	61
Penetration	float64	Numeric	547.0, 403.0, 253.0	44	82.4	120	15–820	320
Viscosity	float64	Numeric	0.81, 1.46, 3.52	144	42.4	94	0.16–10	2.2

shows a data dictionary that provides a comprehensive overview of the variables associated with plastic-modified bitumen. The dataset includes both categorical variables and essential numerical values. An initial data quality check, which is reported in the dictionary, indicated different levels of completeness. While most of the parameters are well-populated. This understanding is vital for maintaining the dependability of subsequent analysis and model construction.

3. Results

The combined data shows that experimental parameters are highly variable (see Figure 1). Plastic incorporation rates were primarily between 0.5% and 12%, with a subset of research looking at concentrations as high as 36%. The mixing temperature was primarily kept between 160 and 180 °C. The most evaluated plastic modifiers were polyethylene variations (LDPE, HDPE, and LLDPE) and PET.

A continuous tendency across multiple tests is an increase in softening point and a decrease in penetration value with the addition of plastic waste, showing that the binder stiffens. For example, adding 8% LDPE improved the softening temperature from 43 °C to 63 °C [4]. Similarly, at a 12% dose, R-LLDPE enhanced the softening point from 44.1 °C to 122.3 °C [10]. This stiffening effect is ideal for boosting asphalt rutting resistance in hot areas. Viscosity often rises with plastic content, improving binder cohesion but boosting mixing and compaction temperatures, which is an important concern in plant manufacturing.

The following

Table 4. Summary statistics by plastic type.

Plastic Type	n (soft)	Mean Soft (°C)	Range Soft (°C)	n (Pen)	Mean Pen (dmm)	Range Pen (dmm)	n (Visc)	Mean visc (Pa.S)	Range Visc (Pa.s)
Crumbed rubber	2	44.75	42–47	2	61	54–68	0	-	-
E-waste plastic	5	67.04	54–85	5	261	113.33–643	5	1.492	0.99–2.010
EBA	1	58	58–58	1	560	560–560	1	0.94	0.94–0.94
EVA	1	57	57–57	1	500	500–500	1	0.98	0.98–0.98
HDPE	17	69.59	54–87	22	245.59	70–470	6	1.023	0.54–1.54
HP	0	-	-	0	-	-	4	1.25	0.5–2.4
Hybrid	8	82.08	67.33–115	8	123	29–250	0	-	-
LDPE	14	62.57	43–77	23	384.22	38–730	4	0.7	0.54–0.84
LDPE and LLDPE	5	61	5–131	0	-	-	0	-	-
LLDPE	1	59	59–59	1	210	210–210	1	0.38	0.38–0.38

summarizes counts, means, and ranges for softening point (°C), penetration (dmm), and viscosity (Pa·s) by plastic type, computed from the curated dataset (excluding missing values).

As shown in

Table 5. Summary of model performance results.

Model	Target	Avg. MAE	Avg. RSME	Avg_R2	Samples
Median Baseline	Softening Point (°C)	14.154	21.4906	−0.0682	193
Linear Model (Ridge)	Softening Point (°C)	14.4171	18.6615	0.1014	193
Random Forest	Softening Point (°C)	11.3263	16.1304	0.3674	193
XGBoost	Softening Point (°C)	11.0965	15.9136	0.3894	193
Median Baseline	Penetration (dmm)	206.7258	231.3153	−0.0081	206
Linear Model (Ridge)	Penetration (dmm)	235.4897	275.0583	−0.9856	206
Random Forest	Penetration (dmm)	238.3021	275.3899	−0.9717	206
XGBoost	Penetration (dmm)	230.3848	272.7678	−0.9346	206
Median Baseline	Viscosity (Pa.S)	1.7106	2.7141	−0.3008	106
Linear Model (Ridge)	Viscosity (Pa.S)	2.557	3.0389	−4.0487	106
Random Forest	Viscosity (Pa.S)	1.7616	2.1424	−0.8741	106
XGBoost	Viscosity (Pa.S)	1.7879	2.2275	−1.3455	106

above, tree-based models (random forest and XGBoost) outperformed linear and baseline models in terms of predicting softening point and viscosity. XGBoost and random forest were the most effective models for softening point and viscosity, respectively. Some models have negative R² values for penetration and viscosity, indicating poor performance compared to a baseline model. This highlights the complexity and noise in the data for these targets. The good softening point prediction performance shows a more consistent link between inputs and outputs.

Figure 2 compares model performance using MAE, RMSE, and R², revealing significant variances between tested methodologies. The tree-based model outperforms the baseline and linear models in capturing complicated relationships in the dataset, with lower MAE and RMSE values and a higher R². This suggests that nonlinear models are more suited to predicting binder behavior when numerous interacting variables are involved.

In Figure 3a, the graph successfully explained about 79% of the variation in softening point, as seen by the predicted versus real plot (R² = 0.747). The mean absolute error (MAE) of 81.501 means that the average difference between projected and actual values is less than one unit, indicating good predictive capability within the evaluated range. Most of the points are closely aligned with the red 1:1 line, suggesting that the model can accurately capture the trend between actual and anticipated values. However, at higher softening point values, there is visible scattering, indicating that the model underestimates actual performance. This shows that, while the model is generally useful, it may have limitations when dealing with extreme values, or that additional contributing elements, such as polymer type, dosage, or dispersion quality, are not adequately accounted for in the existing features.

The penetration test, as shown in Figure 3b, has moderate accuracy, with an R² value of 0.614, as seen in the anticipated versus actual plot. This shows that the model can explain around 61% of the variation in penetration values. The mean absolute error (MAE) of 7.511 is a larger average difference between anticipated and actual values than the softening point model, implying that penetration is more difficult to predict effectively. While many points match the 1:1 line, the scatter pattern indicates significant variance, particularly at lower and higher penetration rates. This dispersion suggests that additional parameters, such as plastic waste type, particle size, and mixing homogeneity, may have a significant impact on penetration.

Furthermore, in Figure 3c, the model effectively gives about 79% of the variability in viscosity, as shown in the predicted versus real graphic (R² = 0.791). The mean absolute error (MAE) of 0.834 demonstrates that the average forecast error is quite tiny, indicating reasonable accuracy, particularly in the low- to mid-viscosity range. The scatter points are closely aligned with the 1:1 line, particularly for lower and mid-range viscosity values, indicating that the model accurately predicts these ranges. Some variance can be seen at higher actual viscosity levels, when the model tends to slightly underestimate performance, but the overall trend is good.

Similar findings were discovered in plastic-waste modification research, such as that adding waste polyethylene reliably boosts the softening point, increases viscosity, and decreases the penetration [49]. Experiments demonstrate that as plastic content increases, viscosity increases, and penetration decreases, often with nonlinear behavior that complicates straightforward modelling [50]. PET-modified bitumen performs similarly, increasing hardness and improving high-temperature responsiveness at the expense of ductility, which adds scatter to softening point or penetration modelling [51]. Furthermore, ML-based studies on polymers or modified binders underscore the importance of nonlinear approaches in capturing complicated interactions, explaining why viscosity (which is more directly related to polymer content) is more predictable than penetration or softening point [52]. Finally, morphological investigations of polymer dispersion show that microstructure and compatibility play a role in variability, particularly in softening point and penetration predictions [53].

A correlation matrix above (as shown in Figure 4) containing numerical variables was examined. Plastic dose correlated with softening point (≈0.37), viscosity (≈0.27), and penetration value (≈−0.28), indicating a stiffening effect. Temperature and time of mixing revealed very modest relationships with final attributes, implying that within the generally reported ranges, plastic type and volume are more relevant factors.

4. Discussion

• Incorporating plastic waste generally increases softening point, decreases penetration, and increases viscosity, confirming the stiffening effect that improves rutting resistance.
• The softening point model demonstrated good predictive performance (R² = 0.79), although slight underestimation occurred at high values.
• The penetration model showed moderate accuracy (R² = 0.61), reflecting its sensitivity to factors such as plastic type, particle size, and mixing uniformity.
• The viscosity model demonstrated good prediction performance (R² = 0.79), especially in the low- to mid-range viscosity values.
• Permutation analyses indicated that dosage and plastic type were the strongest predictors across targets. In our dataset, predictions became more sensitive beyond ~5% dosage: small changes above this threshold produced larger shifts in softening point and viscosity compared to changes below 5%. Partial-dependence inspection showed monotonic stiffening with dose for PE/PET classes, with diminishing returns beyond ~10–12% as mixing/dispersion limits emerge. Mixing temperature/time exhibited weaker effects within the commonly reported ranges (160–180 °C; 30–90 min).
• Tree-based ML algorithms (Random Forest and XGBoost) outperformed linear models, demonstrating the effectiveness of nonlinear approaches in predicting binder behavior.
• Previous studies support that PE, PET, and other plastic wastes improve high-temperature stability but may reduce ductility.

4.1. Compatibility and Dispersion Behavior

A key limitation in plastic modification of bitumen is the inherently low solubility and chemical compatibility of many polymers with the bitumen medium. Poor dispersion can lead to phase separation, storage instability, and reduced long-term performance. The degree of compatibility differs across polymer types and is influenced by factors such as molecular structure, polarity, particle size, and mixing energy.

Although this review focuses primarily on softening point, penetration, and viscosity due to the availability of reported data, compatibility is an essential criterion that should be consistently evaluated in future research. Including solubility-related indicators such as storage stability, morphological characterization, swelling index, or chemical interaction analysis would allow a more holistic assessment of plastic-modified binders.

4.2. Environmental and Economic Context

Beyond technical performance, sustainability aspects are critical for evaluating plastic-modified asphalt. Life-cycle assessment (LCA) studies report that incorporating recycled plastics can reduce greenhouse gas emissions compared to conventional mixes. For example, ref. [17] demonstrated that PET-modified asphalt mixtures achieved measurable CO₂ reductions at the cradle-to-site stage. Similarly, ref. [2] emphasized that using recycled plastics diverts waste from landfills and reduces reliance on virgin bitumen, supporting circular economy goals.

From an economic perspective, cost comparisons indicate potential savings when replacing a fraction of virgin binder with processed waste plastics. Ref. [28] reported that recycled plastic-modified binders can be competitive with conventional polymer-modified asphalt, though additional processing and sorting costs may offset short-term benefits. TRB guidance [5] recommends a comprehensive life-cycle cost analysis (LCCA) to capture long-term advantages such as extended pavement life and reduced maintenance needs.

4.3. Further Research and Direction

Future studies should incorporate larger datasets and a wider range of dosage levels. Microstructural evaluations using microscopy and spectroscopy can provide deeper insight into dispersion quality and polymer–bitumen interactions. Furthermore, life-cycle assessment (LCA) and cost–benefit analysis will be essential to fully assess environmental and economic sustainability. Compatibility-based performance indicators should be routinely included, as polymer–bitumen solubility greatly influences long-term performance and stability.

5. Conclusions

In this report, the efficiency of machine learning (ML) techniques for predicting the engineering features of plastic-modified bitumen was explored. Using a dataset of 251 experimental records from the literature, three regression tasks were built to model binders’ softening point, penetration, and viscosity. To increase model dependability, data from the same research were excluded using grouped K-fold cross-validation. MAE, RMSE, and R² were used to evaluate model performance. Predicted vs. real plots were also reviewed for interpretation. The following is a summary of the conclusions:

• Four models were evaluated: median baseline, linear regression, random forest, and XGBoost. Tree-based models (random forest and XGBoost) outperformed the other models in predicting binder properties, demonstrating their ability to capture nonlinear interaction.
• Softening point predictions show good accuracy (R² = 0.74, MAE = 81.501), with close alignment between predicted and observed values. However, performance prediction errors increased at the upper-range values, indicating the need for additional features to fully capture variability.
• The penetration predictions demonstrated moderate accuracy (R² = 0.614, MAE = 7.511), showing more sensitivity to uncontrolled parameters such as plastic type, particle size, and mixing homogeneity.
• Viscosity predictions also showed good performance (R² = 0.791, MAE = 0.834), particularly in the low- and mid-range viscosity data.
• The results are consistent with existing research that shows that plastic modification generally increases softening point and viscosity while decreasing penetration. This occurs because thermoplastic polymer chains reinforce the bitumen matrix and restrict molecular movement, thereby stiffening the binder and improving rutting resistance at high service temperatures.
• The findings demonstrate that nonlinear machine learning techniques outperform baseline or linear models in predicting binder performance, encouraging their usage in sustainable pavement design research.